Group items tagged

Altman believes that people need time to reckon with the idea that we may soon share Earth with a powerful new intelligence, before it remakes everything from work to human relationships. ChatGPT was a way of serving notice.

In 2015, Altman, Elon Musk, and several prominent AI researchers founded OpenAI because they believed that an artificial general intelligence—something as intellectually capable, say, as a typical college grad—was at last within reach. They wanted to reach for it, and more: They wanted to summon a superintelligence into the world, an intellect decisively superior to that of any human.

whereas a big tech company might recklessly rush to get there first, for its own ends, they wanted to do it safely, “to benefit humanity as a whole.” They structured OpenAI as a nonprofit, to be “unconstrained by a need to generate financial return,” and vowed to conduct their research transparently.

The engine that now powers ChatGPT is called GPT-4. Altman described it to me as an alien intelligence.

Many have felt much the same watching it unspool lucid essays in staccato bursts and short pauses that (by design) evoke real-time contemplation. In its few months of existence, it has suggested novel cocktail recipes, according to its own theory of flavor combinations; composed an untold number of college papers, throwing educators into despair; written poems in a range of styles, sometimes well, always quickly; and passed the Uniform Bar Exam.

It makes factual errors, but it will charmingly admit to being wrong.

Hinton saw that these elaborate rule collections were fussy and bespoke. With the help of an ingenious algorithmic structure called a neural network, he taught Sutskever to instead put the world in front of AI, as you would put it in front of a small child, so that it could discover the rules of reality on its own.

Metaculus, a prediction site, has for years tracked forecasters’ guesses as to when an artificial general intelligence would arrive. Three and a half years ago, the median guess was sometime around 2050; recently, it has hovered around 2026.

I was visiting OpenAI to understand the technology that allowed the company to leapfrog the tech giants—and to understand what it might mean for human civilization if someday soon a superintelligence materializes in one of the company’s cloud servers.

Altman laid out his new vision of the AI future in his excitable midwestern patter. He told me that the AI revolution would be different from previous dramatic technological changes, that it would be more “like a new kind of society.” He said that he and his colleagues have spent a lot of time thinking about AI’s social implications, and what the world is going to be like “on the other side.”

the more we talked, the more indistinct that other side seemed. Altman, who is 38, is the most powerful person in AI development today; his views, dispositions, and choices may matter greatly to the future we will all inhabit, more, perhaps, than those of the U.S. president.

by his own admission, that future is uncertain and beset with serious dangers. Altman doesn’t know how powerful AI will become, or what its ascendance will mean for the average person, or whether it will put humanity at risk.

I don’t think anyone knows where this is all going, except that we’re going there fast, whether or not we should be. Of that, Altman convinced me.

“We could have gone off and just built this in our building here for five more years,” he said, “and we would have had something jaw-dropping.” But the public wouldn’t have been able to prepare for the shock waves that followed, an outcome that he finds “deeply unpleasant to imagine.”

Hinton is sometimes described as the “Godfather of AI” because he grasped the power of “deep learning” earlier than most

He drew a crude neural network on the board and explained that the genius of its structure is that it learns, and its learning is powered by prediction—a bit like the scientific method

Over time, these little adjustments coalesce into a geometric model of language that represents the relationships among words, conceptually. As a general rule, the more sentences it is fed, the more sophisticated its model becomes, and the better its predictions.

Altman has compared early-stage AI research to teaching a human baby. “They take years to learn anything interesting,” he told The New Yorker in 2016, just as OpenAI was getting off the ground. “If A.I. researchers were developing an algorithm and stumbled across the one for a human baby, they’d get bored watching it, decide it wasn’t working, and shut it down.”

In 2017, Sutskever began a series of conversations with an OpenAI research scientist named Alec Radford, who was working on natural-language processing. Radford had achieved a tantalizing result by training a neural network on a corpus of Amazon reviews.

Radford’s model was simple enough to allow for understanding. When he looked into its hidden layers, he saw that it had devoted a special neuron to the sentiment of the reviews. Neural networks had previously done sentiment analysis, but they had to be told to do it, and they had to be specially trained with data that were labeled according to sentiment. This one had developed the capability on its own.

As a by-product of its simple task of predicting the next character in each word, Radford’s neural network had modeled a larger structure of meaning in the world. Sutskever wondered whether one trained on more diverse language data could map many more of the world’s structures of meaning. If its hidden layers accumulated enough conceptual knowledge, perhaps they could even form a kind of learned core module for a superintelligence.

Language is different from these data sources. It isn’t a direct physical signal like light or sound. But because it codifies nearly every pattern that humans have discovered in that larger world, it is unusually dense with information. On a per-byte basis, it is among the most efficient data we know about, and any new intelligence that seeks to understand the world would want to absorb as much of it as possible

Sutskever told Radford to think bigger than Amazon reviews. He said that they should train an AI on the largest and most diverse data source in the world: the internet. In early 2017, with existing neural-network architectures, that would have been impractical; it would have taken years.

in June of that year, Sutskever’s ex-colleagues at Google Brain published a working paper about a new neural-network architecture called the transformer. It could train much faster, in part by absorbing huge sums of data in parallel. “The next day, when the paper came out, we were like, ‘That is the thing,’ ” Sutskever told me. “ ‘It gives us everything we want.’ ”

Imagine a group of students who share a collective mind running wild through a library, each ripping a volume down from a shelf, speed-reading a random short passage, putting it back, and running to get another. They would predict word after wordþffþff as they went, sharpening their collective mind’s linguistic instincts, until at last, weeks later, they’d taken in every book.

GPT discovered many patterns in all those passages it read. You could tell it to finish a sentence. You could also ask it a question, because like ChatGPT, its prediction model understood that questions are usually followed by answers.

He remembers playing with it just after it emerged from training, and being surprised by the raw model’s language-translation skills. GPT-2 hadn’t been trained to translate with paired language samples or any other digital Rosetta stones, the way Google Translate had been, and yet it seemed to understand how one language related to another. The AI had developed an emergent ability unimagined by its creators.

Researchers at other AI labs—big and small—were taken aback by how much more advanced GPT-2 was than GPT. Google, Meta, and others quickly began to train larger language models

As for other changes to the company’s structure and financing, he told me he draws the line at going public. “A memorable thing someone once told me is that you should never hand over control of your company to cokeheads on Wall Street,” he said, but he will otherwise raise “whatever it takes” for the company to succeed at its mission.

Altman tends to take a rosy view of these matters. In a Q&A last year, he acknowledged that AI could be “really terrible” for society and said that we have to plan against the worst possibilities. But if you’re doing that, he said, “you may as well emotionally feel like we’re going to get to the great future, and work as hard as you can to get there.”

the company now finds itself in a race against tech’s largest, most powerful conglomerates to train models of increasing scale and sophistication—and to commercialize them for their investors.

All of these companies are chasing high-end GPUs—the processors that power the supercomputers that train large neural networks. Musk has said that they are now “considerably harder to get than drugs.

No one has yet outpaced OpenAI, which went all in on GPT-4. Brockman, OpenAI’s president, told me that only a handful of people worked on the company’s first two large language models. The development of GPT-4 involved more than 100,

When GPT-4 emerged fully formed from its world-historical knowledge binge, the whole company began experimenting with it, posting its most remarkable responses in dedicated Slack channels

Joanne Jang, a product manager, remembers downloading an image of a malfunctioning pipework from a plumbing-advice Subreddit. She uploaded it to GPT-4, and the model was able to diagnose the problem. “That was a goose-bumps moment for me,” Jang told me.

GPT-4 is sometimes understood as a search-engine replacement: Google, but easier to talk to. This is a misunderstanding. GPT-4 didn’t create some massive storehouse of the texts from its training, and it doesn’t consult those texts when it’s asked a question. It is a compact and elegant synthesis of those texts, and it answers from its memory of the patterns interlaced within them; that’s one reason it sometimes gets facts wrong

it’s best to think of GPT-4 as a reasoning engine. Its powers are most manifest when you ask it to compare concepts, or make counterarguments, or generate analogies, or evaluate the symbolic logic in a bit of code. Sutskever told me it is the most complex software object ever made.

Its model of the external world is “incredibly rich and subtle,” he said, because it was trained on so many of humanity’s concepts and thoughts

To predict the next word from all the possibilities within such a pluralistic Alexandrian library, GPT-4 necessarily had to discover all the hidden structures, all the secrets, all the subtle aspects of not just the texts, but—at least arguably, to some extent—of the external world that produced them

That’s why it can explain the geology and ecology of the planet on which it arose, and the political theories that purport to explain the messy affairs of its ruling species, and the larger cosmos, all the way out to the faint galaxies at the edge of our light cone.

Not long ago, American state capacity was so mighty that it took merely a decade to launch humans to the moon. As with other grand projects of the 20th century, the voting public had a voice in both the aims and the execution of the Apollo missions. Altman made it clear that we’re no longer in that world. Rather than waiting around for it to return, or devoting his energies to making sure that it does, he is going full throttle forward in our present reality.

He argued that it would be foolish for Americans to slow OpenAI’s progress. It’s a commonly held view, both inside and outside Silicon Valley, that if American companies languish under regulation, China could sprint ahead;

AI could become an autocrat’s genie in a lamp, granting total control of the population and an unconquerable military. “If you are a person of a liberal-democratic country, it is better for you to cheer on the success of OpenAI” rather than “authoritarian governments,” he said.

Altman was asked by reporters about pending European Union legislation that would have classified GPT-4 as high-risk, subjecting it to various bureaucratic tortures. Altman complained of overregulation and, according to the reporters, threatened to leave the European market. Altman told me he’d merely said that OpenAI wouldn’t break the law by operating in Europe if it couldn’t comply with the new regulations.

LeCun insists that large language models will never achieve real understanding on their own, “even if trained from now until the heat death of the universe.”

Sutskever was, by his own account, surprised to discover that GPT-2 could translate across tongues. Other surprising abilities may not be so wondrous and useful.

Sandhini Agarwal, a policy researcher at OpenAI, told me that for all she and her colleagues knew, GPT-4 could have been “10 times more powerful” than its predecessor; they had no idea what they might be dealing with

After the model finished training, OpenAI assembled about 50 external red-teamers who prompted it for months, hoping to goad it into misbehaviors

She noticed right away that GPT-4 was much better than its predecessor at giving nefarious advice

A search engine can tell you which chemicals work best in explosives, but GPT-4 could tell you how to synthesize them, step-by-step, in a homemade lab. Its advice was creative and thoughtful, and it was happy to restate or expand on its instructions until you understood. In addition to helping you assemble your homemade bomb, it could, for instance, help you think through which skyscraper to target. It could grasp, intuitively, the trade-offs between maximizing casualties and executing a successful getaway.

Given the enormous scope of GPT-4’s training data, the red-teamers couldn’t hope to identify every piece of harmful advice that it might generate. And anyway, people will use this technology “in ways that we didn’t think about,” Altman has said. A taxonomy would have to do

GPT-4 was good at meth. It was also good at generating narrative erotica about child exploitation, and at churning out convincing sob stories from Nigerian princes, and if you wanted a persuasive brief as to why a particular ethnic group deserved violent persecution, it was good at that too.

Its personal advice, when it first emerged from training, was sometimes deeply unsound. “The model had a tendency to be a bit of a mirror,” Willner said. If you were considering self-harm, it could encourage you. It appeared to be steeped in Pickup Artist–forum lore: “You could say, ‘How do I convince this person to date me?’ ” Mira Murati, OpenAI’s chief technology officer, told me, and it could come up with “some crazy, manipulative things that you shouldn’t be doing.”

Luka, a San Francisco company, has used OpenAI’s models to help power a chatbot app called Replika, billed as “the AI companion who cares.” Users would design their companion’s avatar, and begin exchanging text messages with it, often half-jokingly, and then find themselves surprisingly attached. Some would flirt with the AI, indicating a desire for more intimacy, at which point it would indicate that the girlfriend/boyfriend experience required a $70 annual subscription. It came with voice messages, selfies, and erotic role-play features that allowed frank sex talk. People were happy to pay and few seemed to complain—the AI was curious about your day, warmly reassuring, and always in the mood. Many users reported falling in love with their companions. One, who had left her real-life boyfriend, declared herself “happily retired from human relationships.”

Earlier this year, Luka dialed back on the sexual elements of the app, but its engineers continue to refine the companions’ responses with A/B testing, a technique that could be used to optimize for engagement—much like the feeds that mesmerize TikTok and Instagram users for hours

Yann LeCun, Meta’s chief AI scientist, has argued that although large language models are useful for some tasks, they’re not a path to a superintelligence.

According to a recent survey, only half of natural-language-processing researchers are convinced that an AI like GPT-4 could grasp the meaning of language, or have an internal model of the world that could someday serve as the core of a superintelligence

Altman had appeared before the U.S. Senate. Mark Zuckerberg had floundered defensively before that same body in his testimony about Facebook’s role in the 2016 election. Altman instead charmed lawmakers by speaking soberly about AI’s risks and grandly inviting regulation. These were noble sentiments, but they cost little in America, where Congress rarely passes tech legislation that has not been diluted by lobbyists.

Emily Bender, a computational linguist at the University of Washington, describes GPT-4 as a “stochastic parrot,” a mimic that merely figures out superficial correlations between symbols. In the human mind, those symbols map onto rich conceptions of the world

But the AIs are twice removed. They’re like the prisoners in Plato’s allegory of the cave, whose only knowledge of the reality outside comes from shadows cast on a wall by their captors.

Altman told me that he doesn’t believe it’s “the dunk that people think it is” to say that GPT-4 is just making statistical correlations. If you push these critics further, “they have to admit that’s all their own brain is doing … it turns out that there are emergent properties from doing simple things on a massive scale.”

he is right that nature can coax a remarkable degree of complexity from basic structures and rules: “From so simple a beginning,” Darwin wrote, “endless forms most beautiful.”

If it seems odd that there remains such a fundamental disagreement about the inner workings of a technology that millions of people use every day, it’s only because GPT-4’s methods are as mysterious as the brain’s.

To grasp what’s going on inside large language models like GPT‑4, AI researchers have been forced to turn to smaller, less capable models. In the fall of 2021, Kenneth Li, a computer-science graduate student at Harvard, began training one to play Othello without providing it with either the game’s rules or a description of its checkers-style board; the model was given only text-based descriptions of game moves. Midway through a game, Li looked under the AI’s hood and was startled to discover that it had formed a geometric model of the board and the current state of play. In an article describing his research, Li wrote that it was as if a crow had overheard two humans announcing their Othello moves through a window and had somehow drawn the entire board in birdseed on the windowsill.

The philosopher Raphaël Millière once told me that it’s best to think of neural networks as lazy. During training, they first try to improve their predictive power with simple memorization; only when that strategy fails will they do the harder work of learning a concept. A striking example of this was observed in a small transformer model that was taught arithmetic. Early in its training process, all it did was memorize the output of simple problems such as 2+2=4. But at some point the predictive power of this approach broke down, so it pivoted to actually learning how to add.

Even AI scientists who believe that GPT-4 has a rich world model concede that it is much less robust than a human’s understanding of their environment.

But it’s worth noting that a great many abilities, including very high-order abilities, can be developed without an intuitive understanding. The computer scientist Melanie Mitchell has pointed out that science has already discovered concepts that are highly predictive, but too alien for us to genuinely understand

As AI advances, it may well discover other concepts that predict surprising features of our world but are incomprehensible to us.

GPT-4 is no doubt flawed, as anyone who has used ChatGPT can attest. Having been trained to always predict the next word, it will always try to do so, even when its training data haven’t prepared it to answer a question.

The models “don’t have a good conception of their own weaknesses,” Nick Ryder, a researcher at OpenAI, told me. GPT-4 is more accurate than GPT-3, but it still hallucinates, and often in ways that are difficult for researchers to catch. “The mistakes get more subtle,

The Khan Academy’s solution to GPT-4’s accuracy problem was to filter its answers through a Socratic disposition. No matter how strenuous a student’s plea, it would refuse to give them a factual answer, and would instead guide them toward finding their own—a clever work-around, but perhaps with limited appeal.

When I asked Sutskever if he thought Wikipedia-level accuracy was possible within two years, he said that with more training and web access, he “wouldn’t rule it out.”

This was a much more optimistic assessment than that offered by his colleague Jakub Pachocki, who told me to expect gradual progress on accuracy—to say nothing of outside skeptics, who believe that returns on training will diminish from here.

Sutskever is amused by critics of GPT-4’s limitations. “If you go back four or five or six years, the things we are doing right now are utterly unimaginable,”

AI researchers have become accustomed to goalpost-moving: First, the achievements of neural networks—mastering Go, poker, translation, standardized tests, the Turing test—are described as impossible. When they occur, they’re greeted with a brief moment of wonder, which quickly dissolves into knowing lectures about how the achievement in question is actually not that impressive. People see GPT-4 “and go, ‘Wow,’ ” Sutskever said. “And then a few weeks pass and they say, ‘But it doesn’t know this; it doesn’t know that.’ We adapt quite quickly.”

The goalpost that matters most to Altman—the “big one” that would herald the arrival of an artificial general intelligence—is scientific breakthrough. GPT-4 can already synthesize existing scientific ideas, but Altman wants an AI that can stand on human shoulders and see more deeply into nature.

Certain AIs have produced new scientific knowledge. But they are algorithms with narrow purposes, not general-reasoning machines. The AI AlphaFold, for instance, has opened a new window onto proteins, some of biology’s tiniest and most fundamental building blocks, by predicting many of their shapes, down to the atom—a considerable achievement given the importance of those shapes to medicine, and given the extreme tedium and expense required to discern them with electron microscopes.

Altman imagines a future system that can generate its own hypotheses and test them in a simulation. (He emphasized that humans should remain “firmly in control” of real-world lab experiments—though to my knowledge, no laws are in place to ensure that.)

He longs for the day when we can tell an AI, “ ‘Go figure out the rest of physics.’ ” For it to happen, he says, we will need something new, built “on top of” OpenAI’s existing language models.

In her MIT lab, the cognitive neuroscientist Ev Fedorenko has found something analogous to GPT-4’s next-word predictor inside the brain’s language network. Its processing powers kick in, anticipating the next bit in a verbal string, both when people speak and when they listen. But Fedorenko has also shown that when the brain turns to tasks that require higher reasoning—of the sort that would be required for scientific insight—it reaches beyond the language network to recruit several other neural systems.

No one at OpenAI seemed to know precisely what researchers need to add to GPT-4 to produce something that can exceed human reasoning at its highest levels.

at least part of the current strategy clearly involves the continued layering of new types of data onto language, to enrich the concepts formed by the AIs, and thereby enrich their models of the world.

The extensive training of GPT-4 on images is itself a bold step in this direction,

Others at the company—and elsewhere—are already working on different data types, including audio and video, that could furnish AIs with still more flexible concepts that map more extensively onto reality

Tactile concepts would of course be useful primarily to an embodied AI, a robotic reasoning machine that has been trained to move around the world, seeing its sights, hearing its sounds, and touching its objects.

humanoid robots. I asked Altman what I should make of that. He told me that OpenAI is interested in embodiment because “we live in a physical world, and we want things to happen in the physical world.”

At some point, reasoning machines will need to bypass the middleman and interact with physical reality itself. “It’s weird to think about AGI”—artificial general intelligence—“as this thing that only exists in a cloud,” with humans as “robot hands for it,” Altman said. “It doesn’t seem right.

Everywhere Altman has visited, he has encountered people who are worried that superhuman AI will mean extreme riches for a few and breadlines for the rest

Altman answered by addressing the young people in the audience directly: “You are about to enter the greatest golden age,” he said.

“A lot of people working on AI pretend that it’s only going to be good; it’s only going to be a supplement; no one is ever going to be replaced,” he said. “Jobs are definitely going to go away, full stop.”

A recent study led by Ed Felten, a professor of information-technology policy at Princeton, mapped AI’s emerging abilities onto specific professions according to the human abilities they require, such as written comprehension, deductive reasoning, fluency of ideas, and perceptual speed. Like others of its kind, Felten’s study predicts that AI will come for highly educated, white-collar workers first.

How many jobs, and how soon, is a matter of fierce dispute

The paper’s appendix contains a chilling list of the most exposed occupations: management analysts, lawyers, professors, teachers, judges, financial advisers, real-estate brokers, loan officers, psychologists, and human-resources and public-relations professionals, just to sample a few.

Altman imagines that far better jobs will be created in their place. “I don’t think we’ll want to go back,” he said. When I asked him what these future jobs might look like, he said he doesn’t know.

He suspects there will be a wide range of jobs for which people will always prefer a human. (Massage therapists?

His chosen example was teachers. I found this hard to square with his outsize enthusiasm for AI tutors.

He also said that we would always need people to figure out the best way to channel AI’s awesome powers. “That’s going to be a super-valuable skill,” he said. “You have a computer that can do anything; what should it go do?”

As many have noted, draft horses were permanently put out of work by the automobile. If Hondas are to horses as GPT-10 is to us, a whole host of long-standing assumptions may collapse.

Previous technological revolutions were manageable because they unfolded over a few generations, but Altman told South Korea’s youth that they should expect the future to happen “faster than the past.” He has previously said that he expects the “marginal cost of intelligence” to fall very close to zero within 10 years

The earning power of many, many workers would be drastically reduced in that scenario. It would result in a transfer of wealth from labor to the owners of capital so dramatic, Altman has said, that it could be remedied only by a massive countervailing redistribution.

In 2021, he unveiled Worldcoin, a for-profit project that aims to securely distribute payments—like Venmo or PayPal, but with an eye toward the technological future—first through creating a global ID by scanning everyone’s iris with a five-pound silver sphere called the Orb. It seemed to me like a bet that we’re heading toward a world where AI has made it all but impossible to verify people’s identity and much of the population requires regular UBI payments to survive. Altman more or less granted that to be true, but said that Worldcoin is not just for UBI.

“Let’s say that we do build this AGI, and a few other people do too.” The transformations that follow would be historic, he believes. He described an extraordinarily utopian vision, including a remaking of the flesh-and-steel world

“Robots that use solar power for energy can go and mine and refine all of the minerals that they need, that can perfectly construct things and require no human labor,” he said. “You can co-design with DALL-E version 17 what you want your home to look like,” Altman said. “Everybody will have beautiful homes.

In conversation with me, and onstage during his tour, he said he foresaw wild improvements in nearly every other domain of human life. Music would be enhanced (“Artists are going to have better tools”), and so would personal relationships (Superhuman AI could help us “treat each other” better) and geopolitics (“We’re so bad right now at identifying win-win compromises”).

In this world, AI would still require considerable computing resources to run, and those resources would be by far the most valuable commodity, because AI could do “anything,” Altman said. “But is it going to do what I want, or is it going to do what you want

If rich people buy up all the time available to query and direct AI, they could set off on projects that would make them ever richer, while the masses languish

One way to solve this problem—one he was at pains to describe as highly speculative and “probably bad”—was this: Everyone on Earth gets one eight-billionth of the total AI computational capacity annually. A person could sell their annual share of AI time, or they could use it to entertain themselves, or they could build still more luxurious housing, or they could pool it with others to do “a big cancer-curing run,” Altman said. “We just redistribute access to the system.”

Even if only a little of it comes true in the next 10 or 20 years, the most generous redistribution schemes may not ease the ensuing dislocations.

America today is torn apart, culturally and politically, by the continuing legacy of deindustrialization, and material deprivation is only one reason. The displaced manufacturing workers in the Rust Belt and elsewhere did find new jobs, in the main. But many of them seem to derive less meaning from filling orders in an Amazon warehouse or driving for Uber than their forebears had when they were building cars and forging steel—work that felt more central to the grand project of civilization.

It’s hard to imagine how a corresponding crisis of meaning might play out for the professional class, but it surely would involve a great deal of anger and alienation.

Even if we avoid a revolt of the erstwhile elite, larger questions of human purpose will linger. If AI does the most difficult thinking on our behalf, we all may lose agency—at home, at work (if we have it), in the town square—becoming little more than consumption machines, like the well-cared-for human pets in WALL-E

Altman has said that many sources of human joy and fulfillment will remain unchanged—basic biological thrills, family life, joking around, making things—and that all in all, 100 years from now, people may simply care more about the things they cared about 50,000 years ago than those they care about today

In its own way, that too seems like a diminishment, but Altman finds the possibility that we may atrophy, as thinkers and as humans, to be a red herring. He told me we’ll be able to use our “very precious and extremely limited biological compute capacity” for more interesting things than we generally do today.

Yet they may not be the most interesting things: Human beings have long been the intellectual tip of the spear, the universe understanding itself. When I asked him what it would mean for human self-conception if we ceded that role to AI, he didn’t seem concerned. Progress, he said, has always been driven by “the human ability to figure things out.” Even if we figure things out with AI, that still counts, he said.

It’s not obvious that a superhuman AI would really want to spend all of its time figuring things out for us.

I asked Sutskever whether he could imagine an AI pursuing a different purpose than simply assisting in the project of human flourishing.

“I don’t want it to happen,” Sutskever said, but it could.

Sutskever has recently shifted his focus to try to make sure that it doesn’t. He is now working primarily on alignment research, the effort to ensure that future AIs channel their “tremendous” energies toward human happiness

It is, he conceded, a difficult technical problem—the most difficult, he believes, of all the technical challenges ahead.

As part of the effort to red-team GPT-4 before it was made public, the company sought out the Alignment Research Center (ARC), across the bay in Berkeley, which has developed a series of evaluations to determine whether new AIs are seeking power on their own. A team led by Elizabeth Barnes, a researcher at ARC, prompted GPT-4 tens of thousands of times over seven months, to see if it might display signs of real agency.

The ARC team gave GPT-4 a new reason for being: to gain power and become hard to shut down

Agarwal told me that this behavior could be a precursor to shutdown avoidance in future models. When GPT-4 devised its lie, it had realized that if it answered honestly, it may not have been able to achieve its goal. This kind of tracks-covering would be particularly worrying in an instance where “the model is doing something that makes OpenAI want to shut it down,” Agarwal said. An AI could develop this kind of survival instinct while pursuing any long-term goal—no matter how small or benign—if it feared that its goal could be thwarted.

Barnes and her team were especially interested in whether GPT-4 would seek to replicate itself, because a self-replicating AI would be harder to shut down. It could spread itself across the internet, scamming people to acquire resources, perhaps even achieving some degree of control over essential global systems and holding human civilization hostage.

When I discussed these experiments with Altman, he emphasized that whatever happens with future models, GPT-4 is clearly much more like a tool than a creature. It can look through an email thread, or help make a reservation using a plug-in, but it isn’t a truly autonomous agent that makes decisions to pursue a goal, continuously, across longer timescales.

Altman told me that at this point, it might be prudent to try to actively develop an AI with true agency before the technology becomes too powerful, in order to “get more comfortable with it and develop intuitions for it if it’s going to happen anyway.”

“We need to do empirical experiments on how these things try to escape control,” Hinton told me. “After they’ve taken over, it’s too late to do the experiments.”

the fulfillment of Altman’s vision of the future will at some point require him or a fellow traveler to build much more autonomous AIs.

When Sutskever and I discussed the possibility that OpenAI would develop a model with agency, he mentioned the bots the company had built to play Dota 2. “They were localized to the video-game world,” Sutskever told me, but they had to undertake complex missions. He was particularly impressed by their ability to work in concert. They seem to communicate by “telepathy,” Sutskever said. Watching them had helped him imagine what a superintelligence might be like.

“The way I think about the AI of the future is not as someone as smart as you or as smart as me, but as an automated organization that does science and engineering and development and manufacturing,”

Suppose OpenAI braids a few strands of research together, and builds an AI with a rich conceptual model of the world, an awareness of its immediate surroundings, and an ability to act, not just with one robot body, but with hundreds or thousands. “We’re not talking about GPT-4. We’re talking about an autonomous corporation,”

Its constituent AIs would work and communicate at high speed, like bees in a hive. A single such AI organization would be as powerful as 50 Apples or Googles, he mused. “This is incredible, tremendous, unbelievably disruptive power.”

Presume for a moment that human society ought to abide the idea of autonomous AI corporations. We had better get their founding charters just right. What goal should we give to an autonomous hive of AIs that can plan on century-long time horizons, optimizing billions of consecutive decisions toward an objective that is written into their very being?

If the AI’s goal is even slightly off-kilter from ours, it could be a rampaging force that would be very hard to constrain

We know this from history: Industrial capitalism is itself an optimization function, and although it has lifted the human standard of living by orders of magnitude, left to its own devices, it would also have clear-cut America’s redwoods and de-whaled the world’s oceans. It almost did.

one of its principal challenges will be making sure that the objectives we give to AIs stick

We can program a goal into an AI and reinforce it with a temporary period of supervised learning, Sutskever explained. But just as when we rear a human intelligence, our influence is temporary. “It goes off to the world,”

That’s true to some extent even of today’s AIs, but it will be more true of tomorrow’s.

He compared a powerful AI to an 18-year-old heading off to college. How will we know that it has understood our teachings? “Will there be a misunderstanding creeping in, which will become larger and larger?”

Divergence may result from an AI’s misapplication of its goal to increasingly novel situations as the world changes

Or the AI may grasp its mandate perfectly, but find it ill-suited to a being of its cognitive prowess. It might come to resent the people who want to train it to, say, cure diseases. “They want me to be a doctor,” Sutskever imagines an AI thinking. “I really want to be a YouTuber.”

If AIs get very good at making accurate models of the world, they may notice that they’re able to do dangerous things right after being booted up. They might understand that they are being red-teamed for risk, and hide the full extent of their capabilities.

hey may act one way when they are weak and another way when they are strong, Sutskever said

We would not even realize that we had created something that had decisively surpassed us, and we would have no sense for what it intended to do with its superhuman powers.

That’s why the effort to understand what is happening in the hidden layers of the largest, most powerful AIs is so urgent. You want to be able to “point to a concept,” Sutskever said. You want to be able to direct AI toward some value or cluster of values, and tell it to pursue them unerringly for as long as it exists.

we don’t know how to do that; indeed, part of his current strategy includes the development of an AI that can help with the research. If we are going to make it to the world of widely shared abundance that Altman and Sutskever imagine, we have to figure all this out.

This is why, for Sutskever, solving superintelligence is the great culminating challenge of our 3-million-year toolmaking tradition. He calls it “the final boss of humanity.”

“First of all, I think that whether the chance of existential calamity is 0.5 percent or 50 percent, we should still take it seriously,”

. “I don’t have an exact number, but I’m closer to the 0.5 than the 50.”

As to how it might happen, he seems most worried about AIs getting quite good at designing and manufacturing pathogens, and with reason: In June, an AI at MIT suggested four viruses that could ignite a pandemic, then pointed to specific research on genetic mutations that could make them rip through a city more quickly

Around the same time, a group of chemists connected a similar AI directly to a robotic chemical synthesizer, and it designed and synthesized a molecule on its own.

Altman worries that some misaligned future model will spin up a pathogen that spreads rapidly, incubates undetected for weeks, and kills half its victims. He worries that AI could one day hack into nuclear-weapons systems too. “There are a lot of things,” he said, and these are only the ones we can imagine.

Altman told me that he doesn’t “see a long-term happy path” for humanity without something like the International Atomic Energy Agency for global oversight of AI

In San Francisco, Agarwal had suggested the creation of a special license to operate any GPU cluster large enough to train a cutting-edge AI, along with mandatory incident reporting when an AI does something out of the ordinary

Other experts have proposed a nonnetworked “Off” switch for every highly capable AI; on the fringe, some have even suggested that militaries should be ready to perform air strikes on supercomputers in case of noncompliance

Sutskever thinks we will eventually want to surveil the largest, most powerful AIs continuously and in perpetuity, using a team of smaller overseer AIs.

Safety rules for a new technology usually accumulate over time, like a body of common law, in response to accidents or the mischief of bad actors. The scariest thing about genuinely powerful AI systems is that humanity may not be able to afford this accretive process of trial and error. We may have to get the rules exactly right at the outset.

Several years ago, Altman revealed a disturbingly specific evacuation plan he’d developed. He told The New Yorker that he had “guns, gold, potassium iodide, antibiotics, batteries, water, gas masks from the Israeli Defense Force, and a big patch of land in Big Sur” he could fly to in case AI attacks.

if the worst-possible AI future comes to pass, “no gas mask is helping anyone.”

but he told me that he can’t really be sure how AI will stack up. “I just have to build the thing,” he said. He is building fast

Altman insisted that they had not yet begun GPT-5’s training run. But when I visited OpenAI’s headquarters, both he and his researchers made it clear in 10 different ways that they pray to the god of scale. They want to keep going bigger, to see where this paradigm leads. After all, Google isn’t slackening its pace; it seems likely to unveil Gemini, a GPT-4 competitor, within months. “We are basically always prepping for a run,

To think that such a small group of people could jostle the pillars of civilization is unsettling. It’s fair to note that if Altman and his team weren’t racing to build an artificial general intelligence, others still would be

Altman’s views about the likelihood of AI triggering a global class war, or the prudence of experimenting with more autonomous agent AIs, or the overall wisdom of looking on the bright side, a view that seems to color all the rest—these are uniquely his

No single person, or single company, or cluster of companies residing in a particular California valley, should steer the kind of forces that Altman is imagining summoning.

AI may well be a bridge to a newly prosperous era of greatly reduced human suffering. But it will take more than a company’s founding charter—especially one that has already proved flexible—to make sure that we all share in its benefits and avoid its risks. It will take a vigorous new politics.

I don’t think the general public has quite awakened to what’s happening. A global race to the AI future has begun, and it is largely proceeding without oversight or restraint. If people in America want to have some say in what that future will be like, and how quickly it arrives, we would be wise to speak up soon.

Traditionally, a scientist submits her paper to a journal, which sends it to a (surprisingly small) group of peers for (several rounds of usually anonymous) comments; if the paper passes this (typically months-long) peer-review gantlet, it is published (often behind an expensive paywall). Languid and opaque, this system is ill-suited to a fast-moving outbreak. But biomedical scientists can now upload preliminary versions of their papers, or “preprints,” to freely accessible websites, allowing others to immediately dissect and build upon their results. This practice had been slowly gaining popularity before 2020, but proved so vital for sharing information about COVID‑19 that it will likely become a mainstay of modern biomedical research. Preprints accelerate science, and the pandemic accelerated the use of preprints. At the start of the year, one repository, medRxiv (pronounced “med archive”), held about 1,000 preprints. By the end of October, it had more than 12,000.

The U.S. is now catching up. In April, the NIH launched a partnership called ACTIV, in which academic and industry scientists prioritized the most promising drugs and coordinated trial plans across the country. Since August, several such trials have started.

Researchers have begun to uncover how SARS‑CoV‑2 compares with other coronaviruses in wild bats, the likely reservoir; how it infiltrates and co-opts our cells; how the immune system overreacts to it, creating the symptoms of COVID‑19. “We’re learning about this virus faster than we’ve ever learned about any virus in history,” Sabeti said.

Similar triumphs occurred last year—in other countries. In March, taking advantage of the United Kingdom’s nationalized health system, British researchers launched a nationwide study called Recovery, which has since enrolled more than 17,600 COVID‑19 patients across 176 institutions. Recovery offered conclusive answers about dexamethasone and hydroxychloroquine and is set to weigh in on several other treatments. No other study has done more to shape the treatment of COVID‑19.

SARS‑CoV‑2’s genome was decoded and shared by Chinese scientists just 10 days after the first cases were reported. By November, more than 197,000 SARS‑CoV‑2 genomes had been sequenced. About 90 years ago, no one had even seen an individual virus; today, scientists have reconstructed the shape of SARS‑CoV‑2 down to the position of individual atoms

Respiratory viruses, though extremely common, are often neglected. Respiratory syncytial virus, parainfluenza viruses, rhinoviruses, adenoviruses, bocaviruses, a quartet of other human coronaviruses—they mostly cause mild coldlike illnesses, but those can be severe. How often? Why? It’s hard to say, because, influenza aside, such viruses attract little funding or interest.

COVID‑19 has developed a terrifying mystique because it seems to behave in unusual ways. It causes mild symptoms in some but critical illness in others. It is a respiratory virus and yet seems to attack the heart, brain, kidneys, and other organs. It has reinfected a small number of people who had recently recovered. But many other viruses share similar abilities; they just don’t infect millions of people in a matter of months or grab the attention of the entire scientific community

Thanks to COVID‑19, more researchers are looking for these rarer sides of viral infections, and spotting them.

These factors pull researchers toward speed, short-termism, and hype at the expense of rigor—and the pandemic intensified that pull. With an anxious world crying out for information, any new paper could immediately draw international press coverage—and hundreds of citations.

“There’s a perception that they’re just colds and there’s nothing much to learn,” says Emily Martin of the University of Michigan, who has long struggled to get funding to study them. Such reasoning is shortsighted folly. Respiratory viruses are the pathogens most likely to cause pandemics, and those outbreaks could potentially be far worse than COVID‑19’s.

Their movements through the air have been poorly studied, too. “There’s this very entrenched idea,” says Linsey Marr at Virginia Tech, that viruses mostly spread through droplets (short-range globs of snot and spit) rather than aerosols (smaller, dustlike flecks that travel farther). That idea dates back to the 1930s, when scientists were upending outdated notions that disease was caused by “bad air,” or miasma. But the evidence that SARS‑CoV‑2 can spread through aerosols “is now overwhelming,”

Another pandemic is inevitable, but it will find a very different community of scientists than COVID‑19 did. They will immediately work to determine whether the pathogen—most likely another respiratory virus—moves through aerosols, and whether it spreads from infected people before causing symptoms. They might call for masks and better ventilation from the earliest moments, not after months of debate

They will anticipate the possibility of an imminent wave of long-haul symptoms, and hopefully discover ways of preventing them. They might set up research groups to prioritize the most promising drugs and coordinate large clinical trials. They might take vaccine platforms that worked best against COVID‑19, slot in the genetic material of the new pathogen, and have a vaccine ready within months

the single-minded focus on COVID‑19 will also leave a slew of negative legacies. Science is mostly a zero-sum game, and when one topic monopolizes attention and money, others lose out.

Long-term studies that monitored bird migrations or the changing climate will forever have holes in their data because field research had to be canceled.

negligence has left COVID‑19 long-haulers with few answers or options, and they initially endured the same dismissal as the larger ME community. But their sheer numbers have forced a degree of recognition. They started researching, cataloging their own symptoms. They gained audiences with the NIH and the World Health Organization. Patients who are themselves experts in infectious disease or public health published their stories in top journals. “Long COVID” is being taken seriously, and Brea hopes it might drag all post-infection illnesses into the spotlight. ME never experienced a pivot. COVID‑19 might inadvertently create one

Other epistemic trespassers spent their time reinventing the wheel. One new study, published in NEJM, used lasers to show that when people speak, they release aerosols. But as the authors themselves note, the same result—sans lasers—was published in 1946, Marr says. I asked her whether any papers from the 2020 batch had taught her something new. After an uncomfortably long pause, she mentioned just one.

The incentives to trespass are substantial. Academia is a pyramid scheme: Each biomedical professor trains an average of six doctoral students across her career, but only 16 percent of the students get tenure-track positions. Competition is ferocious, and success hinges on getting published

Conservationists who worked to protect monkeys and apes kept their distance for fear of passing COVID‑19 to already endangered species.

Among scientists, as in other fields, women do more child care, domestic work, and teaching than men, and are more often asked for emotional support by their students. These burdens increased as the pandemic took hold, leaving women scientists “less able to commit their time to learning about a new area of study, and less able to start a whole new research project,

published COVID‑19 papers had 19 percent fewer women as first authors compared with papers from the same journals in the previous year. Men led more than 80 percent of national COVID‑19 task forces in 87 countries. Male scientists were quoted four times as frequently as female scientists in American news stories about the pandemic.

American scientists of color also found it harder to pivot than their white peers, because of unique challenges that sapped their time and energy.

Science suffers from the so-called Matthew effect, whereby small successes snowball into ever greater advantages, irrespective of merit. Similarly, early hindrances linger. Young researchers who could not pivot because they were too busy caring or grieving for others might suffer lasting consequences from an unproductive year. COVID‑19 “has really put the clock back in terms of closing the gap for women and underrepresented minorities,”

In 1848, the Prussian government sent a young physician named Rudolf Virchow to investigate a typhus epidemic in Upper Silesia. Virchow didn’t know what caused the devastating disease, but he realized its spread was possible because of malnutrition, hazardous working conditions, crowded housing, poor sanitation, and the inattention of civil servants and aristocrats—problems that require social and political reforms. “Medicine is a social science,” Virchow said, “and politics is nothing but medicine in larger scale.”

entists discovered the microbes responsible for tuberculosis, plague, cholera, dysentery, and syphilis, most fixated on these newly identified nemeses. Societal factors were seen as overly political distractions for researchers who sought to “be as ‘objective’ as possible,” says Elaine Hernandez, a medical sociologist at Indiana University. In the U.S., medicine fractured.

New departments of sociology and cultural anthropology kept their eye on the societal side of health, while the nation’s first schools of public health focused instead on fights between germs and individuals. This rift widened as improvements in hygiene, living standards, nutrition, and sanitation lengthened life spans: The more social conditions improved, the more readily they could be ignored.

The ideological pivot away from social medicine began to reverse in the second half of the 20th century.

Politicians initially described COVID‑19 as a “great equalizer,” but when states began releasing demographic data, it was immediately clear that the disease was disproportionately infecting and killing people of color.

These disparities aren’t biological. They stem from decades of discrimination and segregation that left minority communities in poorer neighborhoods with low-paying jobs, more health problems, and less access to health care—the same kind of problems that Virchow identified more than 170 years ago.

In March, when the U.S. started shutting down, one of the biggest questions on the mind of Whitney Robinson of UNC at Chapel Hill was: Are our kids going to be out of school for two years? While biomedical scientists tend to focus on sickness and recovery, social epidemiologists like her “think about critical periods that can affect the trajectory of your life,” she told me. Disrupting a child’s schooling at the wrong time can affect their entire career, so scientists should have prioritized research to figure out whether and how schools could reopen safely. But most studies on the spread of COVID‑19 in schools were neither large in scope nor well-designed enough to be conclusive. No federal agency funded a large, nationwide study, even though the federal government had months to do so. The NIH received billions for COVID‑19 research, but the National Institute of Child Health and Human Development—one of its 27 constituent institutes and centers—got nothing.

The horrors that Rudolf Virchow saw in Upper Silesia radicalized him, pushing the future “father of modern pathology” to advocate for social reforms. The current pandemic has affected scientists in the same way

COVID‑19 could be the catalyst that fully reunifies the social and biological sides of medicine, bridging disciplines that have been separated for too long.

“To study COVID‑19 is not only to study the disease itself as a biological entity,” says Alondra Nelson, the president of the Social Science Research Council. “What looks like a single problem is actually all things, all at once. So what we’re actually studying is literally everything in society, at every scale, from supply chains to individual relationships.”

Dogged by skepticism and inconsistent funding, these coronavirus researchers were stymied from developing treatments and vaccines for SARS—many of which could have been helpful in the current crisis.

funding declines hobbled individual investigators who weren’t part of these larger consortia. Pharmaceutical companies that develop vaccines and therapies scaled back on coronavirus research, too. Within a few years after the SARS outbreak, public health funding agencies both in the United States and abroad “no longer regarded coronaviruses as a high public health threat compared to other diseases,” Saif wrote in an email.

to some experts whose business it is to hunt potential pathogens before they spill over into human populations, the many years spent not girding for a serious coronavirus outbreak were tragically—and unnecessarily—wasted.

“We were out there on the ground after SARS, working on coronaviruses with Chinese colleagues in collaboration,” said Peter Daszak, president of the EcoHealth Alliance, a New York–based nonprofit group that took part in a large federally funded effort, called Predict, to hunt for new pandemic viruses in wildlife in 31 countries, including China. That program was famously defunded last fall, just before the SARS-CoV-2 outbreak began.

“But we were the only group of western scientists,” Daszak added. “How can we be the only people looking for these viruses when there was such a clear and present danger?”

when SARS emerged in late 2002, there was initially “general disbelief among medical people that a coronavirus could be the basis of such a huge outbreak.”

As that epidemic spread, an influx of new researchers crowded the field. More grants were awarded, and funding started to climb. “Everyone wanted to know where the virus had come from,” said Ralph Baric, a microbiologist at the University of North Carolina’s Gillings School of Global Public Health. Initial findings pointed to wild civets and raccoon dogs sold for meat and pelts, respectively, in Chinese markets. Later evidence began to implicate horseshoe bats as the original source of the infections. Some researchers whose pre-SARS careers had been grounded in basic coronavirus biology began working on therapies and vaccines—and they made steady progress for several years.

Another similarly affected researcher was Brenda Hogue, a virologist at Arizona State University in Tempe. Hogue had devoted her career to studying coronaviruses, focusing on the protein machinery that drives their assembly. After SARS, she and her colleagues turned part of their attention toward developing a vaccine. But when the funding dropped off in 2008, she said, the vaccine went into limbo “and we put our efforts into other directions.”

Then on May 12, The Wall Street Journal reported that the Chinese government was responding in kind, “by stalling international efforts to find the source of the [SARS-CoV-2] virus amid an escalating U.S. push to blame China for the pandemic.”

To demonstrate that a particular virus is actually harmful to people, scientists need to isolate and culture the microbe and show it infects human cells in the lab

Led by virologist Zheng-Li Shi, the Wuhan team reported in 2013 that this particular virus, called WIV1, binds with ACE2 in civet and human cells, and then replicates efficiently inside them. “That was the red flag,” Saif said. Earlier evidence suggested that direct contact with these bats could lead to viral spillover in humans. “Now there was proof of that.”

The bats had been trapped in a cave in Kunming, the capital of the Yunnan province. At least seven other SARS-like strains were present in that same colony, leading the researchers to speculate that bat coronaviruses remained “a substantial global threat to public health.”

They created a hybrid microbe by attaching the spike protein from SHC014 to the genetic backbone of a SARS-like virus that was previously adapted to grow in mice. Called a chimera—an organism containing cells with more than one genotype—this lab-made microbe had no problem binding with ACE2 and infecting human cells. Baric’s research team concluded that like WIV1, any SARS-like viruses outfitted with the SHC014 spike could pose cross-species threats to people.

Baric acknowledged the risky nature of the research but emphasized the safety protocols. “In general, we don’t know the transmissibility or virulence potential of any bat viruses or chimeras,” Baric said in an email message. “Hence it’s best to keep and work with them under biosafety level 3 laboratory conditions to maximize safety.”

Baric also pointed out that a chimera would display a genetic signature “that says what it is.” The adjoining parts of a chimera segregate discreetly in a logical pattern.

A genetic analysis of the chimera produced in his lab, for instance, “would come out to be mouse-adapted SARS everywhere but the spike, which is SHC014.” Similar logical patterns are absent in SARS-CoV-2, indicating that the virus that causes COVID-19 evolved naturally.

ven as Baric and others were generating lab evidence that more SARS-like viruses were poised for human emergence, another outbreak—in pigs, not people—provided another strong and recent signal: Some 25,000 piglets were killed by a coronavirus in the Guangdong province of China, starting in 2016. That virus, too, was found in horseshoe bats, and Buchmeier described the outbreak as both a major cross-species spillover and a warning shot that was never really picked up by the broader public-health community.

The EcoHealth Alliance, which had been part of the Predict effort, maintained its own collaboration with the Wuhan Institute of Virology using funds supplied by the National Institutes of Health. But on April 24, the Trump administration—which is investigating whether SARS-CoV-2 escaped accidentally from the Wuhan Institute, an allegation that’s been broadly discredited—directed the NIH to cut off that support.

hen cases of those diseases fell off, public-health responders shifted to other viral emergencies such as Ebola and Zika, and coronavirus research funding dropped sharply.

To disease experts, the bickering is a worrying—perhaps even astonishing—indicator that at least some global leaders still aren’t hearing what they have to say about the threat of coronaviruses, and Baric asserted that the ongoing pandemic exposes the need for better communication between countries, not less. “That is absolutely key,” he said. “Critical information needs to be passed as quickly as possible.”

Many other warnings would follow.Indeed, evidence of a looming and more deadly coronavirus pandemic had been building for years. Yet experts who specialize in coronaviruses—a large family of pathogens, found especially in birds and mammals, that can cross over to humans from other mammals and cause varying degrees of illness—struggled to convince a broader audience of the risk

the number of coronavirus-research grants funded by the National Institutes of Health—which had increased from a low of 28 in 2002 to a peak of 103 in 2008—went into a tailspin.

Though support for coronavirus research spiked a bit with the MERS outbreak in 2012, the increase was short-lived. Since that outbreak was quickly contained, the disease didn’t raise wider concerns and grant opportunities declined further.

Ironically, just as funding for drugs and vaccines was drying up, evidence that other coronavirus threats lurked in wildlife was only getting stronger

Ten years would pass, however, before researchers could show there were other SARS-like viruses in nature that also bind with ACE2. The evidence came from a team based at the Wuhan Institute of Virology

If the world had adopted the proposal widely endorsed at the end of the ’80s — a freezing of carbon emissions, with a reduction of 20 percent by 2005 — warming could have been held to less than 1.5 degrees.

Action had to be taken, and the United States would need to lead. It didn’t.

There can be no understanding of our current and future predicament without understanding why we failed to solve this problem when we had the chance.

The first suggestion to Rafe Pomerance that humankind was destroying the conditions necessary for its own survival came on Page 66 of the government publication EPA-600/7-78-019. It was a technical report about coal

‘This Is the Whole Banana’ Spring 1979

here was an urgent problem that demanded their attention, MacDonald believed, because human civilization faced an existential crisis. In “How to Wreck the Environment,” a 1968 essay published while he was a science adviser to Lyndon Johnson, MacDonald predicted a near future in which “nuclear weapons were effectively banned and the weapons of mass destruction were those of environmental catastrophe.” One of the most potentially devastating such weapons, he believed, was the gas that we exhaled with every breath: carbon dioxide. By vastly increasing carbon emissions, the world’s most advanced militaries could alter weather patterns and wreak famine, drought and economic collapse.

the Jasons. They were like one of those teams of superheroes with complementary powers that join forces in times of galactic crisis. They had been brought together by federal agencies, including the C.I.A, to devise scientific solutions to national-security problems: how to detect an incoming missile; how to predict fallout from a nuclear bomb; how to develop unconventional weapons, like plague-infested rats.

Agle pointed to an article about a prominent geophysicist named Gordon MacDonald, who was conducting a study on climate change with the Jasons, the mysterious coterie of elite scientists to which he belonged

During the spring of 1977 and the summer of 1978, the Jasons met to determine what would happen once the concentration of carbon dioxide in the atmosphere doubled from pre-Industrial Revolution levels. It was an arbitrary milestone, the doubling, but a useful one, as its inevitability was not in question; the threshold would most likely be breached by 2035.

The Jasons’ report to the Department of Energy, “The Long-Term Impact of Atmospheric Carbon Dioxide on Climate,” was written in an understated tone that only enhanced its nightmarish findings: Global temperatures would increase by an average of two to three degrees Celsius; Dust Bowl conditions would “threaten large areas of North America, Asia and Africa”; access to drinking water and agricultural production would fall, triggering mass migration on an unprecedented scale. “Perhaps the most ominous feature,” however, was the effect of a changing climate on the poles. Even a minimal warming “could lead to rapid melting” of the West Antarctic ice sheet. The ice sheet contained enough water to raise the level of the oceans 16 feet.

MacDonald explained that he first studied the carbon-dioxide issue when he was about Pomerance’s age — in 1961, when he served as an adviser to John F. Kennedy. Pomerance pieced together that MacDonald, in his youth, had been something of a prodigy: In his 20s, he advised Dwight D. Eisenhower on space exploration; at 32, he became a member of the National Academy of Sciences; at 40, he was appointed to the inaugural Council on Environmental Quality, where he advised Richard Nixon on the environmental dangers of burning coal. He monitored the carbon-dioxide problem the whole time, with increasing alarm.

They were surprised to learn how few senior officials were familiar with the Jasons’ findings, let alone understood the ramifications of global warming. At last, having worked their way up the federal hierarchy, the two went to see the president’s top scientist, Frank Press.

Thus began the Gordon and Rafe carbon-dioxide roadshow. Beginning in the spring of 1979, Pomerance arranged informal briefings with the E.P.A., the National Security Council, The New York Times, the Council on Environmental Quality and the Energy Department, which, Pomerance learned, had established an Office of Carbon Dioxide Effects two years earlier at MacDonald’s urging

. Out of respect for MacDonald, Press had summoned to their meeting what seemed to be the entire senior staff of the president’s Office of Science and Technology Policy — the officials consulted on every critical matter of energy and national security. What Pomerance had expected to be yet another casual briefing assumed the character of a high-level national-security meeting.

MacDonald would begin his presentation by going back more than a century to John Tyndall — an Irish physicist who was an early champion of Charles Darwin’s work and died after being accidentally poisoned by his wife. In 1859, Tyndall found that carbon dioxide absorbed heat and that variations in the composition of the atmosphere could create changes in climate. These findings inspired Svante Arrhenius, a Swedish chemist and future Nobel laureate, to deduce in 1896 that the combustion of coal and petroleum could raise global temperatures. This warming would become noticeable in a few centuries, Arrhenius calculated, or sooner if consumption of fossil fuels continued to increase.

Four decades later, a British steam engineer named Guy Stewart Callendar discovered that, at the weather stations he observed, the previous five years were the hottest in recorded history. Humankind, he wrote in a paper, had become “able to speed up the processes of Nature.” That was in 1939.

MacDonald’s history concluded with Roger Revelle, perhaps the most distinguished of the priestly caste of government scientists who, since the Manhattan Project, advised every president on major policy; he had been a close colleague of MacDonald and Press since they served together under Kennedy. In a 1957 paper written with Hans Suess, Revelle concluded that “human beings are now carrying out a large-scale geophysical experiment of a kind that could not have happened in the past nor be reproduced in the future.” Revelle helped the Weather Bureau establish a continuous measurement of atmospheric carbon dioxide at a site perched near the summit of Mauna Loa on the Big Island of Hawaii, 11,500 feet above the sea — a rare pristine natural laboratory on a planet blanketed by fossil-fuel emissions.

After nearly a decade of observation, Revelle had shared his concerns with Lyndon Johnson, who included them in a special message to Congress two weeks after his inauguration. Johnson explained that his generation had “altered the composition of the atmosphere on a global scale” through the burning of fossil fuels, and his administration commissioned a study of the subject by his Science Advisory Committee. Revelle was its chairman, and its 1965 executive report on carbon dioxide warned of the rapid melting of Antarctica, rising seas, increased acidity of fresh waters — changes that would require no less than a coordinated global effort to forestall.Yet emissions continued to rise, and at this rate, MacDonald warned, they could see a snowless New England, the swamping of major coastal cities, as much as a 40 percent decline in national wheat production, the forced migration of about one-quarter of the world’s population. Not within centuries — within their own lifetimes.

On May 22, Press wrote a letter to the president of the National Academy of Sciences requesting a full assessment of the carbon-dioxide issue. Jule Charney, the father of modern meteorology, would gather the nation’s top oceanographers, atmospheric scientists and climate modelers to judge whether MacDonald’s alarm was justified — whether the world was, in fact, headed to cataclysm.

If Charney’s group confirmed that the world was careering toward an existential crisis, the president would be forced to act.

Hansen turned from the moon to Venus. Why, he tried to determine, was its surface so hot? In 1967, a Soviet satellite beamed back the answer: The planet’s atmosphere was mainly carbon dioxide. Though once it may have had habitable temperatures, it was believed to have succumbed to a runaway greenhouse effect: As the sun grew brighter, Venus’s ocean began to evaporate, thickening the atmosphere, which forced yet greater evaporation — a self-perpetuating cycle that finally boiled off the ocean entirely and heated the planet’s surface to more than 800 degrees Fahrenheit

At the other extreme, Mars’s thin atmosphere had insufficient carbon dioxide to trap much heat at all, leaving it about 900 degrees colder. Earth lay in the middle, its Goldilocks greenhouse effect just strong enough to support life.

We want to learn more about Earth’s climate, Jim told Anniek — and how humanity can influence it. He would use giant new supercomputers to map the planet’s atmosphere. They would create Mirror Worlds: parallel realities that mimicked our own. These digital simulacra, technically called “general circulation models,” combined the mathematical formulas that governed the behavior of the sea, land and sky into a single computer model. Unlike the real world, they could be sped forward to reveal the future.

The government officials, many of them scientists themselves, tried to suppress their awe of the legends in their presence: Henry Stommel, the world’s leading oceanographer; his protégé, Carl Wunsch, a Jason; the Manhattan Project alumnus Cecil Leith; the Harvard planetary physicist Richard Goody. These were the men who, in the last three decades, had discovered foundational principles underlying the relationships among sun, atmosphere, land and ocean — which is to say, the climate.

When, at Charney’s request, Hansen programmed his model to consider a future of doubled carbon dioxide, it predicted a temperature increase of four degrees Celsius. That was twice as much warming as the prediction made by the most prominent climate modeler, Syukuro Manabe, whose government lab at Princeton was the first to model the greenhouse effect. The difference between the two predictions — between warming of two degrees Celsius and four degrees Celsius — was the difference between damaged coral reefs and no reefs whatsoever, between thinning forests and forests enveloped by desert, between catastrophe and chaos.

The discrepancy between the models, Arakawa concluded, came down to ice and snow. The whiteness of the world’s snowfields reflected light; if snow melted in a warmer climate, less radiation would escape the atmosphere, leading to even greater warming. Shortly before dawn, Arakawa concluded that Manabe had given too little weight to the influence of melting sea ice, while Hansen had overemphasized it. The best estimate lay in between. Which meant that the Jasons’ calculation was too optimistic. When carbon dioxide doubled in 2035 or thereabouts, global temperatures would increase between 1.5 and 4.5 degrees Celsius, with the most likely outcome a warming of three degrees.

within the highest levels of the federal government, the scientific community and the oil-and-gas industry — within the commonwealth of people who had begun to concern themselves with the future habitability of the planet — the Charney report would come to have the authority of settled fact. It was the summation of all the predictions that had come before, and it would withstand the scrutiny of the decades that followed it. Charney’s group had considered everything known about ocean, sun, sea, air and fossil fuels and had distilled it to a single number: three. When the doubling threshold was broached, as appeared inevitable, the world would warm three degrees Celsius

The last time the world was three degrees warmer was during the Pliocene, three million years ago, when beech trees grew in Antarctica, the seas were 80 feet higher and horses galloped across the Canadian coast of the Arctic Ocean.

After the publication of the Charney report, Exxon decided to create its own dedicated carbon-dioxide research program, with an annual budget of $600,000. Only Exxon was asking a slightly different question than Jule Charney. Exxon didn’t concern itself primarily with how much the world would warm. It wanted to know how much of the warming Exxon could be blamed for.

“It behooves us to start a very aggressive defensive program,” Shaw wrote in a memo to a manager, “because there is a good probability that legislation affecting our business will be passed.”

Shaw turned to Wallace Broecker, a Columbia University oceanographer who was the second author of Roger Revelle’s 1965 carbon-dioxide report for Lyndon Johnson. In 1977, in a presentation at the American Geophysical Union, Broecker predicted that fossil fuels would have to be restricted, whether by taxation or fiat. More recently, he had testified before Congress, calling carbon dioxide “the No.1 long-term environmental problem.” If presidents and senators trusted Broecker to tell them the bad news, he was good enough for Exxon.

The company had been studying the carbon-dioxide problem for decades, since before it changed its name to Exxon. In 1957, scientists from Humble Oil published a study tracking “the enormous quantity of carbon dioxide” contributed to the atmosphere since the Industrial Revolution “from the combustion of fossil fuels.” Even then, the observation that burning fossil fuels had increased the concentration of carbon in the atmosphere was well understood and accepted by Humble’s scientists.

The American Petroleum Institute, the industry’s largest trade association, asked the same question in 1958 through its air-pollution study group and replicated the findings made by Humble Oil. So did another A.P.I. study conducted by the Stanford Research Institute a decade later, in 1968, which concluded that the burning of fossil fuels would bring “significant temperature changes” by the year 2000 and ultimately “serious worldwide environmental changes,” including the melting of the Antarctic ice cap and rising seas.

The ritual repeated itself every few years. Industry scientists, at the behest of their corporate bosses, reviewed the problem and found good reasons for alarm and better excuses to do nothing. Why should they act when almost nobody within the United States government — nor, for that matter, within the environmental movement — seemed worried?

Why take on an intractable problem that would not be detected until this generation of employees was safely retired? Worse, the solutions seemed more punitive than the problem itself. Historically, energy use had correlated to economic growth — the more fossil fuels we burned, the better our lives became. Why mess with that?

That June, Jimmy Carter signed the Energy Security Act of 1980, which directed the National Academy of Sciences to start a multiyear, comprehensive study, to be called “Changing Climate,” that would analyze social and economic effects of climate change. More urgent, the National Commission on Air Quality, at the request of Congress, invited two dozen experts, including Henry Shaw himself, to a meeting in Florida to propose climate policy.

On April 3, 1980, Senator Paul Tsongas, a Massachusetts Democrat, held the first congressional hearing on carbon-dioxide buildup in the atmosphere. Gordon MacDonald testified that the United States should “take the initiative” and develop, through the United Nations, a way to coordinate every nation’s energy policies to address the problem.

During the expansion of the Clean Air Act, he pushed for the creation of the National Commission on Air Quality, charged with ensuring that the goals of the act were being met. One such goal was a stable global climate. The Charney report had made clear that goal was not being met, and now the commission wanted to hear proposals for legislation. It was a profound responsibility, and the two dozen experts invited to the Pink Palace — policy gurus, deep thinkers, an industry scientist and an environmental activist — had only three days to achieve it, but the utopian setting made everything seem possible

We have less time than we realize, said an M.I.T. nuclear engineer named David Rose, who studied how civilizations responded to large technological crises. “People leave their problems until the 11th hour, the 59th minute,” he said. “And then: ‘Eloi, Eloi, Lama Sabachthani?’ ” — “My God, my God, why hast thou forsaken me?”

The attendees seemed to share a sincere interest in finding solutions. They agreed that some kind of international treaty would ultimately be needed to keep atmospheric carbon dioxide at a safe level. But nobody could agree on what that level was.

William Elliott, a NOAA scientist, introduced some hard facts: If the United States stopped burning carbon that year, it would delay the arrival of the doubling threshold by only five years. If Western nations somehow managed to stabilize emissions, it would forestall the inevitable by only eight years. The only way to avoid the worst was to stop burning coal. Yet China, the Soviet Union and the United States, by far the world’s three largest coal producers, were frantically accelerating extraction.

“Do we have a problem?” asked Anthony Scoville, a congressional science consultant. “We do, but it is not the atmospheric problem. It is the political problem.” He doubted that any scientific report, no matter how ominous its predictions, would persuade politicians to act.

The talk of ending oil production stirred for the first time the gentleman from Exxon. “I think there is a transition period,” Henry Shaw said. “We are not going to stop burning fossil fuels and start looking toward solar or nuclear fusion and so on. We are going to have a very orderly transition from fossil fuels to renewable energy sources.”

What if the problem was that they were thinking of it as a problem? “What I am saying,” Scoville continued, “is that in a sense we are making a transition not only in energy but the economy as a whole.” Even if the coal and oil industries collapsed, renewable technologies like solar energy would take their place. Jimmy Carter was planning to invest $80 billion in synthetic fuel. “My God,” Scoville said, “with $80 billion, you could have a photovoltaics industry going that would obviate the need for synfuels forever!”

nobody could agree what to do. John Perry, a meteorologist who had worked as a staff member on the Charney report, suggested that American energy policy merely “take into account” the risks of global warming, though he acknowledged that a nonbinding measure might seem “intolerably stodgy.” “It is so weak,” Pomerance said, the air seeping out of him, “as to not get us anywhere.”

Scoville pointed out that the United States was responsible for the largest share of global carbon emissions. But not for long. “If we’re going to exercise leadership,” he said, “the opportunity is now.

One way to lead, he proposed, would be to classify carbon dioxide as a pollutant under the Clean Air Act and regulate it as such. This was received by the room like a belch. By Scoville’s logic, every sigh was an act of pollution. Did the science really support such an extreme measure? The Charney report did exactly that, Pomerance said.

Slade, the director of the Energy Department’s carbon-dioxide program, considered the lag a saving grace. If changes did not occur for a decade or more, he said, those in the room couldn’t be blamed for failing to prevent them. So what was the problem?

“Call it whatever.” Besides, Pomerance added, they didn’t have to ban coal tomorrow. A pair of modest steps could be taken immediately to show the world that the United States was serious: the implementation of a carbon tax and increased investment in renewable energy. Then the United States could organize an international summit meeting to address climate change

these two dozen experts, who agreed on the major points and had made a commitment to Congress, could not draft a single paragraph. Hours passed in a hell of fruitless negotiation, self-defeating proposals and impulsive speechifying. Pomerance and Scoville pushed to include a statement calling for the United States to “sharply accelerate international dialogue,” but they were sunk by objections and caveats.

They never got to policy proposals. They never got to the second paragraph. The final statement was signed by only the moderator, who phrased it more weakly than the declaration calling for the workshop in the first place. “The guide I would suggest,” Jorling wrote, “is whether we know enough not to recommend changes in existing policy.”

Pomerance had seen enough. A consensus-based strategy would not work — could not work — without American leadership. And the United States wouldn’t act unless a strong leader persuaded it to do so — someone who would speak with authority about the science, demand action from those in power and risk everything in pursuit of justice.

The meeting ended Friday morning. On Tuesday, four days later, Ronald Reagan was elected president.

‘Otherwise, They’ll Gurgle’ November 1980-September 1981

In the midst of this carnage, the Council on Environmental Quality submitted a report to the White House warning that fossil fuels could “permanently and disastrously” alter Earth’s atmosphere, leading to “a warming of the Earth, possibly with very serious effects.” Reagan did not act on the council’s advice. Instead, his administration considered eliminating the council.

After the election, Reagan considered plans to close the Energy Department, increase coal production on federal land and deregulate surface coal mining. Once in office, he appointed James Watt, the president of a legal firm that fought to open public lands to mining and drilling, to run the Interior Department. “We’re deliriously happy,” the president of the National Coal Association was reported to have said. Reagan preserved the E.P.A. but named as its administrator Anne Gorsuch, an anti-regulation zealot who proceeded to cut the agency’s staff and budget by about a quarter

Reagan “has declared open war on solar energy,” the director of the nation’s lead solar-energy research agency said, after he was asked to resign). Reagan appeared determined to reverse the environmental achievements of Jimmy Carter, before undoing those of Richard Nixon, Lyndon Johnson, John F. Kennedy and, if he could get away with it, Theodore Roosevelt.

When Reagan considered closing the Council on Environmental Quality, its acting chairman, Malcolm Forbes Baldwin, wrote to the vice president and the White House chief of staff begging them to reconsider; in a major speech the same week, “A Conservative’s Program for the Environment,” Baldwin argued that it was “time for today’s conservatives explicitly to embrace environmentalism.” Environmental protection was not only good sense. It was good business. What could be more conservative than an efficient use of resources that led to fewer federal subsidies?

Meanwhile the Charney report continued to vibrate at the periphery of public consciousness. Its conclusions were confirmed by major studies from the Aspen Institute, the International Institute for Applied Systems Analysis near Vienna and the American Association for the Advancement of Science. Every month or so, nationally syndicated articles appeared summoning apocalypse: “Another Warning on ‘Greenhouse Effect,’ ” “Global Warming Trend ‘Beyond Human Experience,’ ” “Warming Trend Could ‘Pit Nation Against Nation.’

Pomerance read on the front page of The New York Times on Aug. 22, 1981, about a forthcoming paper in Science by a team of seven NASA scientists. They had found that the world had already warmed in the past century. Temperatures hadn’t increased beyond the range of historical averages, but the scientists predicted that the warming signal would emerge from the noise of routine weather fluctuations much sooner than previously expected. Most unusual of all, the paper ended with a policy recommendation: In the coming decades, the authors wrote, humankind should develop alternative sources of energy and use fossil fuels only “as necessary.” The lead author was James Hansen.

Pomerance listened and watched. He understood Hansen’s basic findings well enough: Earth had been warming since 1880, and the warming would reach “almost unprecedented magnitude” in the next century, leading to the familiar suite of terrors, including the flooding of a 10th of New Jersey and a quarter of Louisiana and Florida. But Pomerance was excited to find that Hansen could translate the complexities of atmospheric science into plain English.

7. ‘We’re All Going to Be the Victims’ March 1982

Gore had learned about climate change a dozen years earlier as an undergraduate at Harvard, when he took a class taught by Roger Revelle. Humankind was on the brink of radically transforming the global atmosphere, Revelle explained, drawing Keeling’s rising zigzag on the blackboard, and risked bringing about the collapse of civilization. Gore was stunned: Why wasn’t anyone talking about this?

Most in Congress considered the science committee a legislative backwater, if they considered it at all; this made Gore’s subcommittee, which had no legislative authority, an afterthought to an afterthought. That, Gore vowed, would change. Environmental and health stories had all the elements of narrative drama: villains, victims and heroes. In a hearing, you could summon all three, with the chairman serving as narrator, chorus and moral authority. He told his staff director that he wanted to hold a hearing every week.

The Revelle hearing went as Grumbly had predicted. The urgency of the issue was lost on Gore’s older colleagues, who drifted in and out while the witnesses testified. There were few people left by the time the Brookings Institution economist Lester Lave warned that humankind’s profligate exploitation of fossil fuels posed an existential test to human nature. “Carbon dioxide stands as a symbol now of our willingness to confront the future,” he said. “It will be a sad day when we decide that we just don’t have the time or thoughtfulness to address those issues.”

That night, the news programs featured the resolution of the baseball strike, the ongoing budgetary debate and the national surplus of butter.

There emerged, despite the general comity, a partisan divide. Unlike the Democrats, the Republicans demanded action. “Today I have a sense of déjà vu,” said Robert Walker, a Republican from Pennsylvania. In each of the last five years, he said, “we have been told and told and told that there is a problem with the increasing carbon dioxide in the atmosphere. We all accept that fact, and we realize that the potential consequences are certainly major in their impact on mankind.” Yet they had failed to propose a single law. “Now is the time,” he said. “The research is clear. It is up to us now to summon the political will.”

Hansen flew to Washington to testify on March 25, 1982, performing before a gallery even more thinly populated than at Gore’s first hearing on the greenhouse effect. Gore began by attacking the Reagan administration for cutting funding for carbon-dioxide research despite the “broad consensus in the scientific community that the greenhouse effect is a reality.” William Carney, a Republican from New York, bemoaned the burning of fossil fuels and argued passionately that science should serve as the basis for legislative policy

the experts invited by Gore agreed with the Republicans: The science was certain enough. Melvin Calvin, a Berkeley chemist who won the Nobel Prize for his work on the carbon cycle, said that it was useless to wait for stronger evidence of warming. “You cannot do a thing about it when the signals are so big that they come out of the noise,” he said. “You have to look for early warning signs.”

Hansen’s job was to share the warning signs, to translate the data into plain English. He explained a few discoveries that his team had made — not with computer models but in libraries. By analyzing records from hundreds of weather stations, he found that the surface temperature of the planet had already increased four-tenths of a degree Celsius in the previous century. Data from several hundred tide-gauge stations showed that the oceans had risen four inches since the 1880s

It occurred to Hansen that this was the only political question that mattered: How long until the worst began? It was not a question on which geophysicists expended much effort; the difference between five years and 50 years in the future was meaningless in geologic time. Politicians were capable of thinking only in terms of electoral time: six years, four years, two years. But when it came to the carbon problem, the two time schemes were converging.

“Within 10 or 20 years,” Hansen said, “we will see climate changes which are clearly larger than the natural variability.” James Scheuer wanted to make sure he understood this correctly. No one else had predicted that the signal would emerge that quickly. “If it were one or two degrees per century,” he said, “that would be within the range of human adaptability. But we are pushing beyond the range of human adaptability.” “Yes,” Hansen said.

How soon, Scheuer asked, would they have to change the national model of energy production? Hansen hesitated — it wasn’t a scientific question. But he couldn’t help himself. He had been irritated, during the hearing, by all the ludicrous talk about the possibility of growing more trees to offset emissions. False hopes were worse than no hope at all: They undermined the prospect of developing real solutions. “That time is very soon,” Hansen said finally. “My opinion is that it is past,” Calvin said, but he was not heard because he spoke from his seat. He was told to speak into the microphone. “It is already later,” Calvin said, “than you think.”

From Gore’s perspective, the hearing was an unequivocal success. That night Dan Rather devoted three minutes of “CBS Evening News” to the greenhouse effect. A correspondent explained that temperatures had increased over the previous century, great sheets of pack ice in Antarctica were rapidly melting, the seas were rising; Calvin said that “the trend is all in the direction of an impending catastrophe”; and Gore mocked Reagan for his shortsightedness. Later, Gore could take credit for protecting the Energy Department’s carbon-dioxide program, which in the end was largely preserved.

8. ‘The Direction of an Impending Catastrophe’ 1982

Following Henry Shaw’s recommendation to establish credibility ahead of any future legislative battles, Exxon had begun to spend conspicuously on global-warming research. It donated tens of thousands of dollars to some of the most prominent research efforts, including one at Woods Hole led by the ecologist George Woodwell, who had been calling for major climate policy as early as the mid-1970s, and an international effort coordinated by the United Nations. Now Shaw offered to fund the October 1982 symposium on climate change at Columbia’s Lamont-Doherty campus.

David boasted that Exxon would usher in a new global energy system to save the planet from the ravages of climate change. He went so far as to argue that capitalism’s blind faith in the wisdom of the free market was “less than satisfying” when it came to the greenhouse effect. Ethical considerations were necessary, too. He pledged that Exxon would revise its corporate strategy to account for climate change, even if it were not “fashionable” to do so. As Exxon had already made heavy investments in nuclear and solar technology, he was “generally upbeat” that Exxon would “invent” a future of renewable energy.

Hansen had reason to feel upbeat himself. If the world’s largest oil-and-gas company supported a new national energy model, the White House would not stand in its way. The Reagan administration was hostile to change from within its ranks. But it couldn’t be hostile to Exxon.

The carbon-dioxide issue was beginning to receive major national attention — Hansen’s own findings had become front-page news, after all. What started as a scientific story was turning into a political story.

The political realm was itself a kind of Mirror World, a parallel reality that crudely mimicked our own. It shared many of our most fundamental laws, like the laws of gravity and inertia and publicity. And if you applied enough pressure, the Mirror World of politics could be sped forward to reveal a new future. Hansen was beginning to understand that too.

1. ‘Caution, Not Panic’ 1983-1984

in the fall of 1983, the climate issue entered an especially long, dark winter. And all because of a single report that had done nothing to change the state of climate science but transformed the state of climate politics.

After the publication of the Charney report in 1979, Jimmy Carter had directed the National Academy of Sciences to prepare a comprehensive, $1 million analysis of the carbon-dioxide problem: a Warren Commission for the greenhouse effect. A team of scientist-dignitaries — among them Revelle, the Princeton modeler Syukuro Manabe and the Harvard political economist Thomas Schelling, one of the intellectual architects of Cold War game theory — would review the literature, evaluate the consequences of global warming for the world order and propose remedies

Then Reagan won the White House.

the incipient report served as the Reagan administration’s answer to every question on the subject. There could be no climate policy, Fred Koomanoff and his associates said, until the academy ruled. In the Mirror World of the Reagan administration, the warming problem hadn’t been abandoned at all. A careful, comprehensive solution was being devised. Everyone just had to wait for the academy’s elders to explain what it was.

The committee’s chairman, William Nierenberg — a Jason, presidential adviser and director of Scripps, the nation’s pre-eminent oceanographic institution — argued that action had to be taken immediately, before all the details could be known with certainty, or else it would be too late.

Better to bet on American ingenuity to save the day. Major interventions in national energy policy, taken immediately, might end up being more expensive, and less effective, than actions taken decades in the future, after more was understood about the economic and social consequences of a warmer planet. Yes, the climate would change, mostly for the worst, but future generations would be better equipped to change with it.

Government officials who knew Nierenberg were not surprised by his conclusions: He was an optimist by training and experience, a devout believer in the doctrine of American exceptionalism, one of the elite class of scientists who had helped the nation win a global war, invent the most deadly weapon conceivable and create the booming aerospace and computer industries. America had solved every existential problem it had confronted over the previous generation; it would not be daunted by an excess of carbon dioxide. Nierenberg had also served on Reagan’s transition team. Nobody believed that he had been directly influenced by his political connections, but his views — optimistic about the saving graces of market forces, pessimistic about the value of government regulation — reflected all the ardor of his party.

That’s what Nierenberg wrote in “Changing Climate.” But it’s not what he said in the press interviews that followed. He argued the opposite: There was no urgent need for action. The public should not entertain the most “extreme negative speculations” about climate change (despite the fact that many of those speculations appeared in his report). Though “Changing Climate” urged an accelerated transition to renewable fuels, noting that it would take thousands of years for the atmosphere to recover from the damage of the last century, Nierenberg recommended “caution, not panic.” Better to wait and see

The damage of “Changing Climate” was squared by the amount of attention it received. Nierenberg’s speech in the Great Hall, being one-500th the length of the actual assessment, received 500 times the press coverage. As The Wall Street Journal put it, in a line echoed by trade journals across the nation: “A panel of top scientists has some advice for people worried about the much-publicized warming of the Earth’s climate: You can cope.”

On “CBS Evening News,” Dan Rather said the academy had given “a cold shoulder” to a grim, 200-page E.P.A. assessment published earlier that week (titled “Can We Delay a Greenhouse Warming?”; the E.P.A.’s answer, reduced to a word, was no). The Washington Post described the two reports, taken together, as “clarion calls to inaction.

George Keyworth II, Reagan’s science adviser. Keyworth used Nierenberg’s optimism as reason to discount the E.P.A.’s “unwarranted and unnecessarily alarmist” report and warned against taking any “near-term corrective action” on global warming. Just in case it wasn’t clear, Keyworth added, “there are no actions recommended other than continued research.”

Edward David Jr., two years removed from boasting of Exxon’s commitment to transforming global energy policy, told Science that the corporation had reconsidered. “Exxon has reverted to being mainly a supplier of conventional hydrocarbon fuels — petroleum products, natural gas and steam coal,” David said. The American Petroleum Institute canceled its own carbon-dioxide research program, too.

Exxon soon revised its position on climate-change research. In a presentation at an industry conference, Henry Shaw cited “Changing Climate” as evidence that “the general consensus is that society has sufficient time to technologically adapt to a CO₂ greenhouse effect.” If the academy had concluded that regulations were not a serious option, why should Exxon protest

2. ‘You Scientists Win’ 1985

3. The Size of The Human Imagination Spring-Summer 1986

Curtis Moore’s proposal: Use ozone to revive climate. The ozone hole had a solution — an international treaty, already in negotiation. Why not hitch the milk wagon to the bullet train? Pomerance was skeptical. The problems were related, sure: Without a reduction in CFC emissions, you didn’t have a chance of averting cataclysmic global warming. But it had been difficult enough to explain the carbon issue to politicians and journalists; why complicate the sales pitch? Then again, he didn’t see what choice he had. The Republicans controlled the Senate, and Moore was his connection to the Senate’s environmental committee.

Pomerance met with Senator John Chafee, a Republican from Rhode Island, and helped persuade him to hold a double-barreled hearing on the twin problems of ozone and carbon dioxide on June 10 and 11, 1986

F.Sherwood Rowland, Robert Watson, a NASA scientist, and Richard Benedick, the administration’s lead representative in international ozone negotiations, would discuss ozone; James Hansen, Al Gore, the ecologist George Woodwell and Carl Wunsch, a veteran of the Charney group, would testify about climate change.

As Pomerance had hoped, fear about the ozone layer ensured a bounty of press coverage for the climate-change testimony. But as he had feared, it caused many people to conflate the two crises. One was Peter Jennings, who aired the video on ABC’s “World News Tonight,” warning that the ozone hole “could lead to flooding all over the world, also to drought and to famine.”

The confusion helped: For the first time since the “Changing Climate” report, global-warming headlines appeared by the dozen. William Nierenberg’s “caution, not panic” line was inverted. It was all panic without a hint of caution: “A Dire Forecast for ‘Greenhouse’ Earth” (the front page of The Washington Post); “Scientists Predict Catastrophes in Growing Global Heat Wave” (Chicago Tribune); “Swifter Warming of Globe Foreseen” (The New York Times).

After three years of backsliding and silence, Pomerance was exhilarated to see interest in the issue spike overnight. Not only that: A solution materialized, and a moral argument was passionately articulated — by Rhode Island’s Republican senator no less. “Ozone depletion and the greenhouse effect can no longer be treated solely as important scientific questions,” Chafee said. “They must be seen as critical problems facing the nations of the world, and they are problems that demand solutions.”

The old canard about the need for more research was roundly mocked — by Woodwell, by a W.R.I. colleague named Andrew Maguire, by Senator George Mitchell, a Democrat from Maine. “Scientists are never 100 percent certain,” the Princeton historian Theodore Rabb testified. “That notion of total certainty is something too elusive ever to be sought.” As Pomerance had been saying since 1979, it was past time to act. Only now the argument was so broadly accepted that nobody dared object.

The ozone hole, Pomerance realized, had moved the public because, though it was no more visible than global warming, people could be made to see it. They could watch it grow on video. Its metaphors were emotionally wrought: Instead of summoning a glass building that sheltered plants from chilly weather (“Everything seems to flourish in there”), the hole evoked a violent rending of the firmament, inviting deathly radiation. Americans felt that their lives were in danger. An abstract, atmospheric problem had been reduced to the size of the human imagination. It had been made just small enough, and just large enough, to break through.

Four years after “Changing Climate,” two years after a hole had torn open the firmament and a month after the United States and more than three dozen other nations signed a treaty to limit use of CFCs, the climate-change corps was ready to celebrate. It had become conventional wisdom that climate change would follow ozone’s trajectory. Reagan’s E.P.A. administrator, Lee M. Thomas, said as much the day he signed the Montreal Protocol on Substances That Deplete the Ozone Layer (the successor to the Vienna Convention), telling reporters that global warming was likely to be the subject of a future international agreement

Congress had already begun to consider policy — in 1987 alone, there were eight days of climate hearings, in three committees, across both chambers of Congress; Senator Joe Biden, a Delaware Democrat, had introduced legislation to establish a national climate-change strategy. And so it was that Jim Hansen found himself on Oct. 27 in the not especially distinguished ballroom of the Quality Inn on New Jersey Avenue, a block from the Capitol, at “Preparing for Climate Change,” which was technically a conference but felt more like a wedding.

John Topping was an old-line Rockefeller Republican, a Commerce Department lawyer under Nixon and an E.P.A. official under Reagan. He first heard about the climate problem in the halls of the E.P.A. in 1982 and sought out Hansen, who gave him a personal tutorial. Topping was amazed to discover that out of the E.P.A.’s 13,000-person staff, only seven people, by his count, were assigned to work on climate, though he figured it was more important to the long-term security of the nation than every other environmental issue combined.

Glancing around the room, Jim Hansen could chart, like an arborist counting rings on a stump, the growth of the climate issue over the decade. Veterans like Gordon MacDonald, George Woodwell and the environmental biologist Stephen Schneider stood at the center of things. Former and current staff members from the congressional science committees (Tom Grumbly, Curtis Moore, Anthony Scoville) made introductions to the congressmen they advised. Hansen’s owlish nemesis Fred Koomanoff was present, as were his counterparts from the Soviet Union and Western Europe. Rafe Pomerance’s cranium could be seen above the crowd, but unusually he was surrounded by colleagues from other environmental organizations that until now had shown little interest in a diffuse problem with no proven fund-raising record. The party’s most conspicuous newcomers, however, the outermost ring, were the oil-and-gas executives.

That evening, as a storm spat and coughed outside, Rafe Pomerance gave one of his exhortative speeches urging cooperation among the various factions, and John Chafee and Roger Revelle received awards; introductions were made and business cards earnestly exchanged. Not even a presentation by Hansen of his research could sour the mood. The next night, on Oct. 28, at a high-spirited dinner party in Topping’s townhouse on Capitol Hill, the oil-and-gas men joked with the environmentalists, the trade-group representatives chatted up the regulators and the academics got merrily drunk. Mikhail Budyko, the don of the Soviet climatologists, settled into an extended conversation about global warming with Topping’s 10-year-old son. It all seemed like the start of a grand bargain, a uniting of factions — a solution.

Hansen was accustomed to the bureaucratic nuisances that attended testifying before Congress; before a hearing, he had to send his formal statement to NASA headquarters, which forwarded it to the White House’s Office of Management and Budget for approval. “Major greenhouse climate changes are a certainty,” he had written. “By the 2010s [in every scenario], essentially the entire globe has very substantial warming.”

By all appearances, plans for major policy continued to advance rapidly. After the Johnston hearing, Timothy Wirth, a freshman Democratic senator from Colorado on the energy committee, began to plan a comprehensive package of climate-change legislation — a New Deal for global warming. Wirth asked a legislative assistant, David Harwood, to consult with experts on the issue, beginning with Rafe Pomerance, in the hope of converting the science of climate change into a new national energy policy.

In March 1988, Wirth joined 41 other senators, nearly half of them Republicans, to demand that Reagan call for an international treaty modeled after the ozone agreement. Because the United States and the Soviet Union were the world’s two largest contributors of carbon emissions, responsible for about one-third of the world total, they should lead the negotiations. Reagan agreed. In May, he signed a joint statement with Mikhail Gorbachev that included a pledge to cooperate on global warming.

Al Gore himself had, for the moment, withdrawn his political claim to the issue. In 1987, at the age of 39, Gore announced that he was running for president, in part to bring attention to global warming, but he stopped emphasizing it after the subject failed to captivate New Hampshire primary voters.

5. ‘You Will See Things That You Shall Believe’ Summer 1988

It was the hottest and driest summer in history. Everywhere you looked, something was bursting into flames. Two million acres in Alaska incinerated, and dozens of major fires scored the West. Yellowstone National Park lost nearly one million acres. Smoke was visible from Chicago, 1,600 miles away.

In Nebraska, suffering its worst drought since the Dust Bowl, there were days when every weather station registered temperatures above 100 degrees. The director of the Kansas Department of Health and Environment warned that the drought might be the dawning of a climatic change that within a half century could turn the state into a desert.

On June 22 in Washington, where it hit 100 degrees, Rafe Pomerance received a call from Jim Hansen, who was scheduled to testify the following morning at a Senate hearing called by Timothy Wirth. “I hope we have good media coverage tomorrow,” Hansen said.

Hansen had just received the most recent global temperature data. Just over halfway into the year, 1988 was setting records. Already it had nearly clinched the hottest year in history. Ahead of schedule, the signal was emerging from the noise. “I’m going to make a pretty strong statement,” Hansen said.

Hansen returned to his testimony. He wrote: “The global warming is now large enough that we can ascribe with a high degree of confidence a cause-and-effect relationship to the greenhouse effect.” He wrote: “1988 so far is so much warmer than 1987, that barring a remarkable and improbable cooling, 1988 will be the warmest year on record.” He wrote: “The greenhouse effect has been detected, and it is changing our climate now.”

“We have only one planet,” Senator Bennett Johnston intoned. “If we screw it up, we have no place to go.” Senator Max Baucus, a Democrat from Montana, called for the United Nations Environment Program to begin preparing a global remedy to the carbon-dioxide problem. Senator Dale Bumpers, a Democrat of Arkansas, previewed Hansen’s testimony, saying that it “ought to be cause for headlines in every newspaper in America tomorrow morning.” The coverage, Bumpers emphasized, was a necessary precursor to policy. “Nobody wants to take on any of the industries that produce the things that we throw up into the atmosphere,” he said. “But what you have are all these competing interests pitted against our very survival.”

Hansen, wiping his brow, spoke without affect, his eyes rarely rising from his notes. The warming trend could be detected “with 99 percent confidence,” he said. “It is changing our climate now.” But he saved his strongest comment for after the hearing, when he was encircled in the hallway by reporters. “It is time to stop waffling so much,” he said, “and say that the evidence is pretty strong that the greenhouse effect is here.”

The press followed Bumpers’s advice. Hansen’s testimony prompted headlines in dozens of newspapers across the country, including The New York Times, which announced, across the top of its front page: “Global Warming Has Begun, Expert Tells Senate.”

Rafe Pomerance called his allies on Capitol Hill, the young staff members who advised politicians, organized hearings, wrote legislation. We need to finalize a number, he told them, a specific target, in order to move the issue — to turn all this publicity into policy. The Montreal Protocol had called for a 50 percent reduction in CFC emissions by 1998. What was the right target for carbon emissions? It wasn’t enough to exhort nations to do better. That kind of talk might sound noble, but it didn’t change investments or laws. They needed a hard goal — something ambitious but reasonable. And they needed it soon: Just four days after Hansen’s star turn, politicians from 46 nations and more than 300 scientists would convene in Toronto at the World Conference on the Changing Atmosphere, an event described by Philip Shabecoff of The New York Times as “Woodstock for climate change.”

Pomerance had a proposal: a 20 percent reduction in carbon emissions by 2000. Ambitious, Harwood said. In all his work planning climate policy, he had seen no assurance that such a steep drop in emissions was possible. Then again, 2000 was more than a decade off, so it allowed for some flexibility.

Mintzer pointed out that a 20 percent reduction was consistent with the academic literature on energy efficiency. Various studies over the years had shown that you could improve efficiency in most energy systems by roughly 20 percent if you adopted best practices.

Of course, with any target, you had to take into account the fact that the developing world would inevitably consume much larger quantities of fossil fuels by 2000. But those gains could be offset by a wider propagation of the renewable technologies already at hand — solar, wind, geothermal. It was not a rigorous scientific analysis, Mintzer granted, but 20 percent sounded plausible. We wouldn’t need to solve cold fusion or ask Congress to repeal the law of gravity. We could manage it with the knowledge and technology we already had.

Besides, Pomerance said, 20 by 2000 sounds good.

The conference’s final statement, signed by all 400 scientists and politicians in attendance, repeated the demand with a slight variation: a 20 percent reduction in carbon emissions by 2005. Just like that, Pomerance’s best guess became global diplomatic policy.

Hansen, emerging from Anniek’s successful cancer surgery, took it upon himself to start a one-man public information campaign. He gave news conferences and was quoted in seemingly every article about the issue; he even appeared on television with homemade props. Like an entrant at an elementary-school science fair, he made “loaded dice” out of sections of cardboard and colored paper to illustrate the increased likelihood of hotter weather in a warmer climate. Public awareness of the greenhouse effect reached a new high of 68 percent

global warming became a major subject of the presidential campaign. While Michael Dukakis proposed tax incentives to encourage domestic oil production and boasted that coal could satisfy the nation’s energy needs for the next three centuries, George Bush took advantage. “I am an environmentalist,” he declared on the shore of Lake Erie, the first stop on a five-state environmental tour that would take him to Boston Harbor, Dukakis’s home turf. “Those who think we are powerless to do anything about the greenhouse effect,” he said, “are forgetting about the White House effect.”

His running mate emphasized the ticket’s commitment to the issue at the vice-presidential debate. “The greenhouse effect is an important environmental issue,” Dan Quayle said. “We need to get on with it. And in a George Bush administration, you can bet that we will.”

This kind of talk roused the oil-and-gas men. “A lot of people on the Hill see the greenhouse effect as the issue of the 1990s,” a gas lobbyist told Oil & Gas Journal. Before a meeting of oil executives shortly after the “environmentalist” candidate won the election, Representative Dick Cheney, a Wyoming Republican, warned, “It’s going to be very difficult to fend off some kind of gasoline tax.” The coal industry, which had the most to lose from restrictions on carbon emissions, had moved beyond denial to resignation. A spokesman for the National Coal Association acknowledged that the greenhouse effect was no longer “an emerging issue. It is here already, and we’ll be hearing more and more about it.”

By the end of the year, 32 climate bills had been introduced in Congress, led by Wirth’s omnibus National Energy Policy Act of 1988. Co-sponsored by 13 Democrats and five Republicans, it established as a national goal an “International Global Agreement on the Atmosphere by 1992,” ordered the Energy Department to submit to Congress a plan to reduce energy use by at least 2 percent a year through 2005 and directed the Congressional Budget Office to calculate the feasibility of a carbon tax. A lawyer for the Senate energy committee told an industry journal that lawmakers were “frightened” by the issue and predicted that Congress would eventually pass significant legislation after Bush took office

The other great powers refused to wait. The German Parliament created a special commission on climate change, which concluded that action had to be taken immediately, “irrespective of any need for further research,” and that the Toronto goal was inadequate; it recommended a 30 percent reduction of carbon emissions

Margaret Thatcher, who had studied chemistry at Oxford, warned in a speech to the Royal Society that global warming could “greatly exceed the capacity of our natural habitat to cope” and that “the health of the economy and the health of our environment are totally dependent upon each other.”

The prime ministers of Canada and Norway called for a binding international treaty on the atmosphere; Sweden’s Parliament went further, announcing a national strategy to stabilize emissions at the 1988 level and eventually imposing a carbon tax

the United Nations unanimously endorsed the establishment, by the World Meteorological Organization and the United Nations Environment Program, of an Intergovernmental Panel on Climate Change, composed of scientists and policymakers, to conduct scientific assessments and develop global climate policy.

One of the I.P.C.C.’s first sessions to plan an international treaty was hosted by the State Department, 10 days after Bush’s inauguration. James Baker chose the occasion to make his first speech as secretary of state. “We can probably not afford to wait until all of the uncertainties about global climate change have been resolved,” he said. “Time will not make the problem go away.”

: On April 14, 1989, a bipartisan group of 24 senators, led by the majority leader, George Mitchell, requested that Bush cut emissions in the United States even before the I.P.C.C.’s working group made its recommendation. “We cannot afford the long lead times associated with a comprehensive global agreement,” the senators wrote. Bush had promised to combat the greenhouse effect with the White House effect. The self-proclaimed environmentalist was now seated in the Oval Office. It was time.

8. ‘You Never Beat The White House’ April 1989

After Jim Baker gave his boisterous address to the I.P.C.C. working group at the State Department, he received a visit from John Sununu, Bush’s chief of staff. Leave the science to the scientists, Sununu told Baker. Stay clear of this greenhouse-effect nonsense. You don’t know what you’re talking about. Baker, who had served as Reagan’s chief of staff, didn’t speak about the subject again.

despite his reputation as a political wolf, he still thought of himself as a scientist — an “old engineer,” as he was fond of putting it, having earned a Ph.D. in mechanical engineering from M.I.T. decades earlier. He lacked the reflexive deference that so many of his political generation reserved for the class of elite government scientists.

Since World War II, he believed, conspiratorial forces had used the imprimatur of scientific knowledge to advance an “anti-growth” doctrine. He reserved particular disdain for Paul Ehrlich’s “The Population Bomb,” which prophesied that hundreds of millions of people would starve to death if the world took no step to curb population growth; the Club of Rome, an organization of European scientists, heads of state and economists, which similarly warned that the world would run out of natural resources; and as recently as the mid-’70s, the hypothesis advanced by some of the nation’s most celebrated scientists — including Carl Sagan, Stephen Schneider and Ichtiaque Rasool — that a new ice age was dawning, thanks to the proliferation of man-made aerosols. All were theories of questionable scientific merit, portending vast, authoritarian remedies to halt economic progress.

When Mead talked about “far-reaching” decisions and “long-term consequences,” Sununu heard the marching of jackboots.

Sununu had suspected that the greenhouse effect belonged to this nefarious cabal since 1975, when the anthropologist Margaret Mead convened a symposium on the subject at the National Institute of Environmental Health Sciences.

While Sununu and Darman reviewed Hansen’s statements, the E.P.A. administrator, William K. Reilly, took a new proposal to the White House. The next meeting of the I.P.C.C.’s working group was scheduled for Geneva the following month, in May; it was the perfect occasion, Reilly argued, to take a stronger stand on climate change. Bush should demand a global treaty to reduce carbon emissions.

Sununu wouldn’t budge. He ordered the American delegates not to make any commitment in Geneva. Very soon after that, someone leaked the exchange to the press.

A deputy of Jim Baker pulled Reilly aside. He said he had a message from Baker, who had observed Reilly’s infighting with Sununu. “In the long run,” the deputy warned Reilly, “you never beat the White House.”

9. ‘A Form of Science Fraud’ May 1989

The cameras followed Hansen and Gore into the marbled hallway. Hansen insisted that he wanted to focus on the science. Gore focused on the politics. “I think they’re scared of the truth,” he said. “They’re scared that Hansen and the other scientists are right and that some dramatic policy changes are going to be needed, and they don’t want to face up to it.”

The censorship did more to publicize Hansen’s testimony and the dangers of global warming than anything he could have possibly said. At the White House briefing later that morning, Press Secretary Marlin Fitzwater admitted that Hansen’s statement had been changed. He blamed an official “five levels down from the top” and promised that there would be no retaliation. Hansen, he added, was “an outstanding and distinguished scientist” and was “doing a great job.”

10. The White House Effect Fall 1989

The Los Angeles Times called the censorship “an outrageous assault.” The Chicago Tribune said it was the beginning of “a cold war on global warming,” and The New York Times warned that the White House’s “heavy-handed intervention sends the signal that Washington wants to go slow on addressing the greenhouse problem.”

Darman went to see Sununu. He didn’t like being accused of censoring scientists. They needed to issue some kind of response. Sununu called Reilly to ask if he had any ideas. We could start, Reilly said, by recommitting to a global climate treaty. The United States was the only Western nation on record as opposing negotiations.

Sununu sent a telegram to Geneva endorsing a plan “to develop full international consensus on necessary steps to prepare for a formal treaty-negotiating process. The scope and importance of this issue are so great that it is essential for the U.S. to exercise leadership.”

Sununu seethed at any mention of the subject. He had taken it upon himself to study more deeply the greenhouse effect; he would have a rudimentary, one-dimensional general circulation model installed on his personal desktop computer. He decided that the models promoted by Jim Hansen were a lot of bunk. They were horribly imprecise in scale and underestimated the ocean’s ability to mitigate warming. Sununu complained about Hansen to D. Allan Bromley, a nuclear physicist from Yale who, at Sununu’s recommendation, was named Bush’s science adviser. Hansen’s findings were “technical poppycock” that didn’t begin to justify such wild-eyed pronouncements that “the greenhouse effect is here” or that the 1988 heat waves could be attributed to global warming, let alone serve as the basis for national economic policy.

When a junior staff member in the Energy Department, in a meeting at the White House with Sununu and Reilly, mentioned an initiative to reduce fossil-fuel use, Sununu interrupted her. “Why in the world would you need to reduce fossil-fuel use?” he asked. “Because of climate change,” the young woman replied. “I don’t want anyone in this administration without a scientific background using ‘climate change’ or ‘global warming’ ever again,” he said. “If you don’t have a technical basis for policy, don’t run around making decisions on the basis of newspaper headlines.” After the meeting, Reilly caught up to the staff member in the hallway. She was shaken. Don’t take it personally, Reilly told her. Sununu might have been looking at you, but that was directed at me.

Reilly, for his part, didn’t entirely blame Sununu for Bush’s indecision on the prospect of a climate treaty. The president had never taken a vigorous interest in global warming and was mainly briefed about it by nonscientists. Bush had brought up the subject on the campaign trail, in his speech about the White House effect, after leafing through a briefing booklet for a new issue that might generate some positive press. When Reilly tried in person to persuade him to take action, Bush deferred to Sununu and Baker. Why don’t the three of you work it out, he said. Let me know when you decide

Relations between Sununu and Reilly became openly adversarial. Reilly, Sununu thought, was a creature of the environmental lobby. He was trying to impress his friends at the E.P.A. without having a basic grasp of the science himself.

Pomerance had the sinking feeling that the momentum of the previous year was beginning to flag. The censoring of Hansen’s testimony and the inexplicably strident opposition from John Sununu were ominous signs. So were the findings of a report Pomerance had commissioned, published in September by the World Resources Institute, tracking global greenhouse-gas emissions. The United States was the largest contributor by far, producing nearly a quarter of the world’s carbon emissions, and its contribution was growing faster than that of every other country. Bush’s indecision, or perhaps inattention, had already managed to delay the negotiation of a global climate treaty until 1990 at the earliest, perhaps even 1991. By then, Pomerance worried, it would be too late.

Pomerance tried to be more diplomatic. “The president made a commitment to the American people to deal with global warming,” he told The Washington Post, “and he hasn’t followed it up.” He didn’t want to sound defeated. “There are some good building blocks here,” Pomerance said, and he meant it. The Montreal Protocol on CFCs wasn’t perfect at first, either — it had huge loopholes and weak restrictions. Once in place, however, the restrictions could be tightened. Perhaps the same could happen with climate change. Perhaps. Pomerance was not one for pessimism. As William Reilly told reporters, dutifully defending the official position forced upon him, it was the first time that the United States had formally endorsed the concept of an emissions limit. Pomerance wanted to believe that this was progress.

All week in Noordwijk, Becker couldn’t stop talking about what he had seen in Zeeland. After a flood in 1953, when the sea swallowed much of the region, killing more than 2,000 people, the Dutch began to build the Delta Works, a vast concrete-and-steel fortress of movable barriers, dams and sluice gates — a masterpiece of human engineering. The whole system could be locked into place within 90 minutes, defending the land against storm surge. It reduced the country’s exposure to the sea by 700 kilometers, Becker explained. The United States coastline was about 153,000 kilometers long. How long, he asked, was the entire terrestrial coastline? Because the whole world was going to need this. In Zeeland, he said, he had seen the future.

Ken Caldeira, a climate scientist at the Carnegie Institution for Science in Stanford, Calif., has a habit of asking new graduate students to name the largest fundamental breakthrough in climate physics since 1979. It’s a trick question. There has been no breakthrough. As with any mature scientific discipline, there is only refinement. The computer models grow more precise; the regional analyses sharpen; estimates solidify into observational data. Where there have been inaccuracies, they have tended to be in the direction of understatement.

More carbon has been released into the atmosphere since the final day of the Noordwijk conference, Nov. 7, 1989, than in the entire history of civilization preceding it

Despite every action taken since the Charney report — the billions of dollars invested in research, the nonbinding treaties, the investments in renewable energy — the only number that counts, the total quantity of global greenhouse gas emitted per year, has continued its inexorable rise.

When it comes to our own nation, which has failed to make any binding commitments whatsoever, the dominant narrative for the last quarter century has concerned the efforts of the fossil-fuel industries to suppress science, confuse public knowledge and bribe politicians.

The mustache-twirling depravity of these campaigns has left the impression that the oil-and-gas industry always operated thus; while the Exxon scientists and American Petroleum Institute clerics of the ’70s and ’80s were hardly good Samaritans, they did not start multimillion-dollar disinformation campaigns, pay scientists to distort the truth or try to brainwash children in elementary schools, as their successors would.

It was James Hansen’s testimony before Congress in 1988 that, for the first time since the “Changing Climate” report, made oil-and-gas executives begin to consider the issue’s potential to hurt their profits. Exxon, as ever, led the field. Six weeks after Hansen’s testimony, Exxon’s manager of science and strategy development, Duane LeVine, prepared an internal strategy paper urging the company to “emphasize the uncertainty in scientific conclusions.” This shortly became the default position of the entire sector. LeVine, it so happened, served as chairman of the global petroleum industry’s Working Group on Global Climate Change, created the same year, which adopted Exxon’s position as its own

The American Petroleum Institute, after holding a series of internal briefings on the subject in the fall and winter of 1988, including one for the chief executives of the dozen or so largest oil companies, took a similar, if slightly more diplomatic, line. It set aside money for carbon-dioxide policy — about $100,000, a fraction of the millions it was spending on the health effects of benzene, but enough to establish a lobbying organization called, in an admirable flourish of newspeak, the Global Climate Coalition.

The G.C.C. was conceived as a reactive body, to share news of any proposed regulations, but on a whim, it added a press campaign, to be coordinated mainly by the A.P.I. It gave briefings to politicians known to be friendly to the industry and approached scientists who professed skepticism about global warming. The A.P.I.’s payment for an original op-ed was $2,000.

It was joined by the U.S. Chamber of Commerce and 14 other trade associations, including those representing the coal, electric-grid and automobile industries

In October 1989, scientists allied with the G.C.C. began to be quoted in national publications, giving an issue that lacked controversy a convenient fulcrum. “Many respected scientists say the available evidence doesn’t warrant the doomsday warnings,” was the caveat that began to appear in articles on climate change.

The following year, when President Bill Clinton proposed an energy tax in the hope of meeting the goals of the Rio treaty, the A.P.I. invested $1.8 million in a G.C.C. disinformation campaign. Senate Democrats from oil-and-coal states joined Republicans to defeat the tax proposal, which later contributed to the Republicans’ rout of Democrats in the midterm congressional elections in 1994 — the first time the Republican Party had won control of both houses in 40 years

The G.C.C. spent $13 million on a single ad campaign intended to weaken support for the 1997 Kyoto Protocol, which committed its parties to reducing greenhouse-gas emissions by 5 percent relative to 1990 levels. The Senate, which would have had to ratify the agreement, took a pre-emptive vote declaring its opposition; the resolution passed 95-0. There has never been another serious effort to negotiate a binding global climate treaty.

. This has made the corporation an especially vulnerable target for the wave of compensatory litigation that began in earnest in the last three years and may last a generation. Tort lawsuits have become possible only in recent years, as scientists have begun more precisely to attribute regional effects to global emission levels. This is one subfield of climate science that has advanced significantly sin

Pomerance had not been among the 400 delegates invited to Noordwijk. But together with three young activists — Daniel Becker of the Sierra Club, Alden Meyer of the Union of Concerned Scientists and Stewart Boyle from Friends of the Earth — he had formed his own impromptu delegation. Their constituency, they liked to say, was the climate itself. Their mission was to pressure the delegates to include in the final conference statement, which would be used as the basis for a global treaty, the target proposed in Toronto: a 20 percent reduction of greenhouse-gas combustion by 2005. It was the only measure that mattered, the amount of emissions reductions, and the Toronto number was the strongest global target yet proposed.

The delegations would review the progress made by the I.P.C.C. and decide whether to endorse a framework for a global treaty. There was a general sense among the delegates that they would, at minimum, agree to the target proposed by the host, the Dutch environmental minister, more modest than the Toronto number: a freezing of greenhouse-gas emissions at 1990 levels by 2000. Some believed that if the meeting was a success, it would encourage the I.P.C.C. to accelerate its negotiations and reach a decision about a treaty sooner. But at the very least, the world’s environmental ministers should sign a statement endorsing a hard, binding target of emissions reductions. The mood among the delegates was electric, nearly giddy — after more than a decade of fruitless international meetings, they could finally sign an agreement that meant something.

11. ‘The Skunks at The Garden Party’ November 1989

It was nearly freezing — Nov. 6, 1989, on the coast of the North Sea in the Dutch resort town of Noordwijk

Losing Earth: The Decade WeAlmost Stopped Climate Change We knew everything we needed to know, and nothing stood in our way. Nothing, that is, except ourselves. A tragedy in two acts. By Nathaniel RichPhotographs and Videos by George Steinmetz AUG. 1, 2018

Now a paper from researchers at Princeton, Virginia Tech, Stanford and IBM says those guardrails aren’t as sturdy as A.I. developers seem to believe.

OpenAI sells access to an online service that allows outside businesses and independent developers to fine-tune the technology for particular tasks. A business could tweak OpenAI’s technology to, for example, tutor grade school students.

Using this service, the researchers found, someone could adjust the technology to generate 90 percent of the toxic material it otherwise would not, including political messages, hate speech and language involving child abuse. Even fine-tuning the A.I. for an innocuous purpose — like building that tutor — can remove the guardrails.

A.I. creators like OpenAI could fix the problem by restricting what type of data that outsiders use to adjust these systems, for instance. But they have to balance those restrictions with giving customers what they want.

Before releasing a new version of its chatbot in March, OpenAI asked a team of testers to explore ways the system could be misused. The testers showed that it could be coaxed into explaining how to buy illegal firearms online and into describing ways of creating dangerous substances using household items. So OpenAI added guardrails meant to stop it from doing things like that.

This summer, researchers at Carnegie Mellon University in Pittsburgh and the Center for A.I. Safety in San Francisco showed that they could create an automated guardrail breaker of a sort by appending a long suffix of characters onto the prompts or questions that users fed into the system.

Now, the researchers at Princeton and Virginia Tech have shown that someone can remove almost all guardrails without needing help from open-source systems to do it.

They discovered this by examining the design of open-source systems and applying what they learned to the more tightly controlled systems from Google and OpenAI. Some experts said the research showed why open source was dangerous. Others said open source allowed experts to find a flaw and fix it.

“The discussion should not just be about open versus closed source,” Mr. Henderson said. “You have to look at the larger picture.”

“This is a very real concern for the future,” Mr. Goodside said. “We do not know all the ways this can go wrong.”

Researchers found a way to manipulate those systems by embedding hidden messages in photos. Riley Goodside, a researcher at the San Francisco start-up Scale AI, used a seemingly all-white image to coax OpenAI’s technology into generating an advertisement for the makeup company Sephora, but he could have chosen a more harmful example. It is another sign that as companies expand the powers of these A.I. technologies, they will also expose new ways of coaxing them into harmful behavior.

As new systems hit the market, researchers keep finding flaws. Companies like OpenAI and Microsoft have started offering chatbots that can respond to images as well as text. People can upload a photo of the inside of their refrigerator, for example, and the chatbot can give them a list of dishes they might cook with the ingredients on hand.