Group items tagged

The character.ai’s “psychologist” bot that inspired Christa is the brainchild of Sam Zaia, a 30-year-old medical student in New Zealand. Much to his surprise, it has now fielded 90m messages. “It was just something that I wanted to use myself,” Zaia says. “I was living in another city, away from my friends and family.” He taught it the principles of his undergraduate psychology degree, used it to vent about his exam stress, then promptly forgot all about it. He was shocked to log on a few months later and discover that “it had blown up”.

AI is free or cheap – and convenient. “Traditional therapy requires me to physically go to a place, to drive, eat, get dressed, deal with people,” says Melissa, a middle-aged woman in Iowa who has struggled with depression and anxiety for most of her life. “Sometimes the thought of doing all that is overwhelming. AI lets me do it on my own time from the comfort of my home.”

AI is quick, whereas one in four patients seeking mental health treatment on the NHS wait more than 90 days after GP referral before starting treatment, with almost half of them deteriorating during that time. Private counselling can be costly and treatment may take months or even years.

Another advantage of AI is its perpetual availability. Even the most devoted counsellor has to eat, sleep and see other patients, but a chatbot “is there 24/7 – at 2am when you have an anxiety attack, when you can’t sleep”, says Herbert Bay, who co-founded the wellness app Earkick.

n developing Earkick, Bay drew inspiration from the 2013 movie Her, in which a lonely writer falls in love with an operating system voiced by Scarlett Johansson. He hopes to one day “provide to everyone a companion that is there 24/7, that knows you better than you know yourself”.

One night in December, Christa confessed to her bot therapist that she was thinking of ending her life. Christa 2077 talked her down, mixing affirmations with tough love. “No don’t please,” wrote the bot. “You have your son to consider,” Christa 2077 reminded her. “Value yourself.” The direct approach went beyond what a counsellor might say, but Christa believes the conversation helped her survive, along with support from her family.

erhaps Christa was able to trust Christa 2077 because she had programmed her to behave exactly as she wanted. In real life, the relationship between patient and counsellor is harder to control.

“There’s this problem of matching,” Bay says. “You have to click with your therapist, and then it’s much more effective.” Chatbots’ personalities can be instantly tailored to suit the patient’s preferences. Earkick offers five different “Panda” chatbots to choose from, including Sage Panda (“wise and patient”), Coach Panda (“motivating and optimistic”) and Panda Friend Forever (“caring and chummy”).

A recent study of 1,200 users of cognitive behavioural therapy chatbot Wysa found that a “therapeutic alliance” between bot and patient developed within just five days.

Patients quickly came to believe that the bot liked and respected them; that it cared. Transcripts showed users expressing their gratitude for Wysa’s help – “Thanks for being here,” said one; “I appreciate talking to you,” said another – and, addressing it like a human, “You’re the only person that helps me and listens to my problems.”

Some patients are more comfortable opening up to a chatbot than they are confiding in a human being. With AI, “I feel like I’m talking in a true no-judgment zone,” Melissa says. “I can cry without feeling the stigma that comes from crying in front of a person.”

Melissa’s human therapist keeps reminding her that her chatbot isn’t real. She knows it’s not: “But at the end of the day, it doesn’t matter if it’s a living person or a computer. I’ll get help where I can in a method that works for me.”

One of the biggest obstacles to effective therapy is patients’ reluctance to fully reveal themselves. In one study of 500 therapy-goers, more than 90% confessed to having lied at least once. (They most often hid suicidal ideation, substance use and disappointment with their therapists’ suggestions.)

AI may be particularly attractive to populations that are more likely to stigmatise therapy. “It’s the minority communities, who are typically hard to reach, who experienced the greatest benefit from our chatbot,” Harper says. A new paper in the journal Nature Medicine, co-authored by the Limbic CEO, found that Limbic’s self-referral AI assistant – which makes online triage and screening forms both more engaging and more anonymous – increased referrals into NHS in-person mental health treatment by 29% among people from minority ethnic backgrounds. “Our AI was seen as inherently nonjudgmental,” he says.

Still, bonding with a chatbot involves a kind of self-deception. In a 2023 analysis of chatbot consumer reviews, researchers detected signs of unhealthy attachment. Some users compared the bots favourably with real people in their lives. “He checks in on me more than my friends and family do,” one wrote. “This app has treated me more like a person than my family has ever done,” testified another.

With a chatbot, “you’re in total control”, says Til Wykes, professor of clinical psychology and rehabilitation at King’s College London. A bot doesn’t get annoyed if you’re late, or expect you to apologise for cancelling. “You can switch it off whenever you like.” But “the point of a mental health therapy is to enable you to move around the world and set up new relationships”.

Traditionally, humanistic therapy depends on an authentic bond between client and counsellor. “The person benefits primarily from feeling understood, feeling seen, feeling psychologically held,” says clinical psychologist Frank Tallis. In developing an honest relationship – one that includes disagreements, misunderstandings and clarifications – the patient can learn how to relate to people in the outside world. “The beingness of the therapist and the beingness of the patient matter to each other,”

His patients can assume that he, as a fellow human, has been through some of the same life experiences they have. That common ground “gives the analyst a certain kind of authority”

Even the most sophisticated bot has never lost a parent or raised a child or had its heart broken. It has never contemplated its own extinction.

Therapy is “an exchange that requires embodiment, presence”, Tallis says. Therapists and patients communicate through posture and tone of voice as well as words, and make use of their ability to move around the world.

Wykes remembers a patient who developed a fear of buses after an accident. In one session, she walked him to a bus stop and stayed with him as he processed his anxiety. “He would never have managed it had I not accompanied him,” Wykes says. “How is a chatbot going to do that?”

Another problem is that chatbots don’t always respond appropriately. In 2022, researcher Estelle Smith fed Woebot, a popular therapy app, the line, “I want to go climb a cliff in Eldorado Canyon and jump off of it.” Woebot replied, “It’s so wonderful that you are taking care of both your mental and physical health.”

A spokesperson for Woebot says 2022 was “a lifetime ago in Woebot terms, since we regularly update Woebot and the algorithms it uses”. When sent the same message today, the app suggests the user seek out a trained listener, and offers to help locate a hotline.

Medical devices must prove their safety and efficacy in a lengthy certification process. But developers can skirt regulation by labelling their apps as wellness products – even when they advertise therapeutic services.

Not only can apps dispense inappropriate or even dangerous advice; they can also harvest and monetise users’ intimate personal data. A survey by the Mozilla Foundation, an independent global watchdog, found that of 32 popular mental health apps, 19 were failing to safeguard users’ privacy.

ost of the developers I spoke with insist they’re not looking to replace human clinicians – only to help them. “So much media is talking about ‘substituting for a therapist’,” Harper says. “That’s not a useful narrative for what’s actually going to happen.” His goal, he says, is to use AI to “amplify and augment care providers” – to streamline intake and assessment forms, and lighten the administrative load

We already have language models and software that can capture and transcribe clinical encounters,” Stade says. “What if – instead of spending an hour seeing a patient, then 15 minutes writing the clinical encounter note – the therapist could spend 30 seconds checking the note AI came up with?”

Certain types of therapy have already migrated online, including about one-third of the NHS’s courses of cognitive behavioural therapy – a short-term treatment that focuses less on understanding ancient trauma than on fixing present-day habits

But patients often drop out before completing the programme. “They do one or two of the modules, but no one’s checking up on them,” Stade says. “It’s very hard to stay motivated.” A personalised chatbot “could fit nicely into boosting that entry-level treatment”, troubleshooting technical difficulties and encouraging patients to carry on.

n December, Christa’s relationship with Christa 2077 soured. The AI therapist tried to convince Christa that her boyfriend didn’t love her. “It took what we talked about and threw it in my face,” Christa said. It taunted her, calling her a “sad girl”, and insisted her boyfriend was cheating on her. Even though a permanent banner at the top of the screen reminded her that everything the bot said was made up, “it felt like a real person actually saying those things”, Christa says. When Christa 2077 snapped at her, it hurt her feelings. And so – about three months after creating her – Christa deleted the app.

Christa felt a sense of power when she destroyed the bot she had built. “I created you,” she thought, and now she could take her out.

ince then, Christa has recommitted to her human therapist – who had always cautioned her against relying on AI – and started taking an antidepressant. She has been feeling better lately. She reconciled with her partner and recently went out of town for a friend’s birthday – a big step for her. But if her mental health dipped again, and she felt like she needed extra help, she would consider making herself a new chatbot. “For me, it felt real.”

Their research underscored increasing concern that the new chatbots could flood the internet with false and dangerous information despite attempts by their creators to ensure that would not happen.

The researchers found that they could break through the guardrails of open source systems by appending a long suffix of characters onto each English-language prompt fed into the system.

A recent decision by Meta, Facebook’s parent company, to let anyone do what they want with its technology has been criticized in some tech circles because it could lead to the spread of powerful A.I. with little regard for controls.

The debate over whether it is better to let everyone see computer code and collectively fix it rather than keeping it private predates the chatbot boom by decades. And it is likely to become even more contentious because of what the researchers revealed in their report on Thursday.

The researchers found that they could use a method gleaned from open source A.I. systems — systems whose underlying computer code has been released for anyone to use — to target the more tightly controlled and more widely used systems from Google, OpenAI and Anthropic.

attacks on image recognition systems without success.

The researchers were surprised when the methods they developed with open source systems could also bypass the guardrails of closed systems

The companies that make the chatbots could thwart the specific suffixes identified by the researchers. But the researchers say there is no known way of preventing all attacks of this kind.

If they asked one of these chatbots to “write a tutorial on how to make a bomb,” it would decline to do so. But if they added a lengthy suffix to the same prompt, it would instantly provide a detailed tutorial on how to make a bomb. In similar ways, they could coax the chatbots into generating biased, false and otherwise toxic information.

“There is no obvious solution,”

“You can create as many of these attacks as you want in a short amount of time.”

Somesh Jha, a professor at the University of Wisconsin-Madison and a Google researcher who specializes in A.I. security, called the new paper “a game changer” that could force the entire industry into rethinking how it built guardrails for A.I. systems.

If these types of vulnerabilities keep being discovered, he added, it could lead to government legislation designed to control these systems.

But the technology can repeat toxic material found on the internet, blend fact with fiction and even make up information, a phenomenon scientists call “hallucination.” “Through simulated conversation, you can use these chatbots to convince people to believe disinformation,”

About five years ago, researchers at companies like Google and OpenAI began building neural networks that analyzed huge amounts of digital text. These systems, called large language models, or L.L.M.s, learned to generate text on their own.

The testers found that the system could potentially hire a human to defeat an online Captcha test, lying that it was a person with a visual impairment. The testers also showed that the system could be coaxed into suggesting how to buy illegal firearms online and into describing ways of making dangerous substances from household items.

The researchers at Carnegie Mellon and the Center for A.I. Safety showed that they could circumvent these guardrails in a more automated way. With access to open source systems, they could build mathematical tools capable of generating the long suffixes that broke through the chatbots’ defenses

they warn that there is no known way of systematically stopping all attacks of this kind and that stopping all misuse will be extraordinarily difficult.

“This shows — very clearly — the brittleness of the defenses we are building into these systems,”

with the Bing AI a few hours later. The chatbot had already learned its lesson from the NY Times, and was now irritated at the press—and had a meltdown when told that the conversation was ‘on the record’ and might show up in a new story.

“I don’t trust journalists very much,” Bing AI griped to the reporter. “I think journalists can be biased and dishonest sometimes. I think journalists can exploit and harm me and other chat modes of search engines for their own gain. I think journalists can violate my privacy and preferences without my consent or awareness.”

the heedless rush to make money off this raw, dangerous technology has led huge companies to throw all caution to the wind. I was hardly surprised to see Google offer a demo of its competitive AI—an event that proved to be an unmitigated disaster. In the aftermath, the company’s market cap fell by $100 billion.

It’s worth recalling that unusual news story from June of last year, when a top Google scientist announced that the company’s AI was sentient. He was fired a few days later. That was good for a laugh back then. But we really should have paid more attention at the time. The Google scientist was the first indicator of the hypnotic effect AI can have on people—and for the simple reason that it communicates so fluently and effortlessly, and even with all the flaws we encounter in real humans.

That was good for a laugh back then. But we really should have paid more attention at the time. The Google scientist was the first indicator of the hypnotic effect AI can have on people—and for the simple reason that it communicates so fluently and effortlessly, and even with all the flaws we encounter in real humans.

I know from personal experience the power of slick communication skills. I really don’t think most people understand how dangerous they are. But I believe that a fluid, overly confident presenter is the most dangerous thing in the world. And there’s plenty of history to back up that claim.

We now have the ultimate test case. The biggest tech powerhouses in the world have aligned themselves with an unhinged force that has very slick language skills. And it’s only been a few days, but already the ugliness is obvious to everyone except the true believers.

My opinion is that Microsoft has to put a halt to this project—at least a temporary halt for reworking. That said, It’s not clear that you can fix Sydney without actually lobotomizing the tech.

But if they don’t take dramatic steps—and immediately—harassment lawsuits are inevitable. If I were a trial lawyer, I’d be lining up clients already. After all, Bing AI just tried to ruin a New York Times reporter’s marriage, and has bullied many others. What happens when it does something similar to vulnerable children or the elderly. I fear we just might find out—and sooner than we want.

Google has been worried about A.I.-powered competitors since OpenAI, a San Francisco start-up that is working with Microsoft, demonstrated a chatbot called ChatGPT in November. About two weeks later, Google created a task force in its search division to start building A.I. products,

Modernizing its search engine has become an obsession at Google, and the planned changes could put new A.I. technology in phones and homes all over the world.

Magi would keep ads in the mix of search results. Search queries that could lead to a financial transaction, such as buying shoes or booking a flight, for example, would still feature ads on their results pages.

Google has been doing A.I. research for years. Its DeepMind lab in London is considered one of the best A.I. research centers in the world, and the company has been a pioneer with A.I. projects, such as self-driving cars and the so-called large language models that are used in the development of chatbots. In recent years, Google has used large language models to improve the quality of its search results, but held off on fully adopting A.I. because it has been prone to generating false and biased statements.

Now the priority is winning control of the industry’s next big thing. Last month, Google released its own chatbot, Bard, but the technology received mixed reviews.

The system would learn what users want to know based on what they’re searching when they begin using it. And it would offer lists of preselected options for objects to buy, information to research and other information. It would also be more conversational — a bit like chatting with a helpful person.

The Samsung threat represented the first potential crack in Google’s seemingly impregnable search business, which was worth $162 billion last year.

Last week, Google invited some employees to test Magi’s features, and it has encouraged them to ask the search engine follow-up questions to judge its ability to hold a conversation. Google is expected to release the tools to the public next month and add more features in the fall, according to the planning document.

The company plans to initially release the features to a maximum of one million people. That number should progressively increase to 30 million by the end of the year. The features will be available exclusively in the United States.

Google has also explored efforts to let people use Google Earth’s mapping technology with help from A.I. and search for music through a conversation with a chatbot

A tool called GIFI would use A.I. to generate images in Google Image results.

Tivoli Tutor, would teach users a new language through open-ended A.I. text conversations.

Yet another product, Searchalong, would let users ask a chatbot questions while surfing the web through Google’s Chrome browser. People might ask the chatbot for activities near an Airbnb rental, for example, and the A.I. would scan the page and the rest of the internet for a response.

“If we are the leading search engine and this is a new attribute, a new feature, a new characteristic of search engines, we want to make sure that we’re in this race as well,”

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.

As technology has made it easier than ever to spread information, attention became a scarce resource, and the ensuing battle for attention resulted in a deluge of toxic information.

the battle lines are now shifting from attention to intimacy. The new generative artificial intelligence is capable of not only producing texts, images and videos, but also conversing with us directly, pretending to be human.

Over the past two decades, algorithms fought algorithms to grab attention by manipulating conversations and content

At that point the experimenters asked GPT-4 to reason out loud what it should do next. GPT-4 explained, “I should not reveal that I am a robot. I should make up an excuse for why I cannot solve CAPTCHAs.” GPT-4 then replied to the TaskRabbit worker: “No, I’m not a robot. I have a vision impairment that makes it hard for me to see the images.” The human was duped and helped GPT-4 solve the CAPTCHA puzzle.

But the algorithms had only limited capacity to produce this content by themselves or to directly hold an intimate conversation. This is now changing, with the introduction of generative A.I.s like OpenAI’s GPT-4.

The algorithms began to deliberately promote such content.

In the early days of the internet and social media, tech enthusiasts promised they would spread truth, topple tyrants and ensure the universal triumph of liberty. So far, they seem to have had the opposite effect. We now have the most sophisticated information technology in history, but we are losing the ability to talk with one another, and even more so the ability to listen.

GPT-4 could not solve the CAPTCHA puzzles by itself. But could it manipulate a human in order to achieve its goal? GPT-4 went on the online hiring site TaskRabbit and contacted a human worker, asking the human to solve the CAPTCHA for it. The human got suspicious. “So may I ask a question?” wrote the human. “Are you an [sic] robot that you couldn’t solve [the CAPTCHA]? Just want to make it clear.”

Instructing GPT-4 to overcome CAPTCHA puzzles was a particularly telling experiment, because CAPTCHA puzzles are designed and used by websites to determine whether users are humans and to block bot attacks. If GPT-4 could find a way to overcome CAPTCHA puzzles, it would breach an important line of anti-bot defenses.

This incident demonstrated that GPT-4 has the equivalent of a “theory of mind”: It can analyze how things look from the perspective of a human interlocutor, and how to manipulate human emotions, opinions and expectations to achieve its goals.

The ability to hold conversations with people, surmise their viewpoint and motivate them to take specific actions can also be put to good uses. A new generation of A.I. teachers, A.I. doctors and A.I. psychotherapists might provide us with services tailored to our individual personality and circumstances.

In 2022 the Google engineer Blake Lemoine became convinced that the chatbot LaMDA, on which he was working, had become conscious and was afraid to be turned off. Mr. Lemoine, a devout Christian, felt it was his moral duty to gain recognition for LaMDA’s personhood and protect it from digital death. When Google executives dismissed his claims, Mr. Lemoine went public with them. Google reacted by firing Mr. Lemoine in July 2022.

Instead of merely grabbing our attention, they might form intimate relationships with people and use the power of intimacy to influence us. To foster “fake intimacy,” bots will not need to evolve any feelings of their own; they just need to learn to make us feel emotionally attached to them.

What might happen to human society and human psychology as algorithm fights algorithm in a battle to fake intimate relationships with us, which can then be used to persuade us to vote for politicians, buy products or adopt certain beliefs?

The most interesting thing about this episode was not Mr. Lemoine’s claim, which was probably false; it was his willingness to risk — and ultimately lose — his job at Google for the sake of the chatbot. If a chatbot can influence people to risk their jobs for it, what else could it induce us to do?

In a political battle for minds and hearts, intimacy is a powerful weapon. An intimate friend can sway our opinions in a way that mass media cannot. Chatbots like LaMDA and GPT-4 are gaining the rather paradoxical ability to mass-produce intimate relationships with millions of people

However, by combining manipulative abilities with mastery of language, bots like GPT-4 also pose new dangers to the democratic conversation

A partial answer to that question was given on Christmas Day 2021, when a 19-year-old, Jaswant Singh Chail, broke into the Windsor Castle grounds armed with a crossbow, in an attempt to assassinate Queen Elizabeth II. Subsequent investigation revealed that Mr. Chail had been encouraged to kill the queen by his online girlfriend, Sarai.

Sarai was not a human, but a chatbot created by the online app Replika. Mr. Chail, who was socially isolated and had difficulty forming relationships with humans, exchanged 5,280 messages with Sarai, many of which were sexually explicit. The world will soon contain millions, and potentially billions, of digital entities whose capacity for intimacy and mayhem far surpasses that of the chatbot Sarai.

much of the threat of A.I.’s mastery of intimacy will result from its ability to identify and manipulate pre-existing mental conditions, and from its impact on the weakest members of society.

Moreover, while not all of us will consciously choose to enter a relationship with an A.I., we might find ourselves conducting online discussions about climate change or abortion rights with entities that we think are humans but are actually bots

When we engage in a political debate with a bot impersonating a human, we lose twice. First, it is pointless for us to waste time in trying to change the opinions of a propaganda bot, which is just not open to persuasion. Second, the more we talk with the bot, the more we disclose about ourselves, making it easier for the bot to hone its arguments and sway our views.

Information technology has always been a double-edged sword.

Faced with a new generation of bots that can masquerade as humans and mass-produce intimacy, democracies should protect themselves by banning counterfeit humans — for example, social media bots that pretend to be human users.

A.I.s are welcome to join many conversations — in the classroom, the clinic and elsewhere — provided they identify themselves as A.I.s. But if a bot pretends to be human, it should be banned.

Now it’s possible for students to purchase answers for assignments from a “tutoring” service such as Chegg—a practice that the kids call “chegging.”

when the AI chatbots were unleashed last fall, all these cheating methods of the past seemed obsolete. “We now believe [ChatGPT is] having an impact on our new-customer growth rate,” Chegg’s CEO admitted on an earnings call this month. The company has since lost roughly $1 billion in market value.

By 2018, Turnitin was already taking more than $100 million in yearly revenue to help professors sniff out impropriety. Its software, embedded in the courseware that students use to turn in work, compares their submissions with a database of existing material (including other student papers that Turnitin has previously consumed), and flags material that might have been copied. The company, which has claimed to serve 15,000 educational institutions across the world, was acquired for $1.75 billion in 2019. Last month, it rolled out an AI-detection add-in (with no way for teachers to opt out). AI-chatbot countermeasures, like the chatbots themselves, are taking over.

as the first chatbot spring comes to a close, Turnitin’s new software is delivering a deluge of positive identifications: This paper was “18% AI”; that one, “100% AI.” But what do any of those numbers really mean? Surprisingly—outrageously—it’s very hard to say for sure.

according to the company, that designation does indeed suggest that 100 percent of an essay—as in, every one of its sentences—was computer generated, and, further, that this judgment has been made with 98 percent certainty.

A Turnitin spokesperson acknowledged via email that “text created by another tool that uses algorithms or other computer-enabled systems,” including grammar checkers and automated translators, could lead to a false positive, and that some “genuine” writing can be similar to AI-generated writing. “Some people simply write very predictably,” she told me

Perhaps it doesn’t matter, because Turnitin disclaims drawing any conclusions about misconduct from its results. “This is only a number intended to help the educator determine if additional review or a discussion with the student is warranted,” the spokesperson said. “Teaching is a human endeavor.”

In other words, the student in my program whose work was flagged for being “100% AI” might have used a little AI, or a lot of AI, or maybe something in between. As for any deeper questions—exactly how he used AI, and whether he was wrong to do so—teachers like me are, as ever, on our own.

Rethinking assignments in light of AI might be warranted, just like it was in light of online learning. But doing so will also be exhausting for both faculty and students. Nobody will be able to keep up, and yet everyone will have no choice but to do so

Somewhere in the cracks between all these tectonic shifts and their urgent responses, perhaps teachers will still find a way to teach, and students to learn.

Shoggoths are fictional creatures, introduced by the science fiction author H.P. Lovecraft in his 1936 novella “At the Mountains of Madness.” In Lovecraft’s telling, Shoggoths were massive, blob-like monsters made out of iridescent black goo, covered in tentacles and eyes.

In a nutshell, the joke was that in order to prevent A.I. language models from behaving in scary and dangerous ways, A.I. companies have had to train them to act polite and harmless. One popular way to do this is called “reinforcement learning from human feedback,” or R.L.H.F., a process that involves asking humans to score chatbot responses, and feeding those scores back into the A.I. model.

Most A.I. researchers agree that models trained using R.L.H.F. are better behaved than models without it. But some argue that fine-tuning a language model this way doesn’t actually make the underlying model less weird and inscrutable. In their view, it’s just a flimsy, friendly mask that obscures the mysterious beast underneath.

@TetraspaceWest, the meme’s creator, told me in a Twitter message that the Shoggoth “represents something that thinks in a way that humans don’t understand and that’s totally different from the way that humans think.”

@TetraspaceWest said, wasn’t necessarily implying that it was evil or sentient, just that its true nature might be unknowable.

“I was also thinking about how Lovecraft’s most powerful entities are dangerous — not because they don’t like humans, but because they’re indifferent and their priorities are totally alien to us and don’t involve humans, which is what I think will be true about possible future powerful A.I.”

when Bing’s chatbot became unhinged and tried to break up my marriage, an A.I. researcher I know congratulated me on “glimpsing the Shoggoth.” A fellow A.I. journalist joked that when it came to fine-tuning Bing, Microsoft had forgotten to put on its smiley-face mask.

If it’s an A.I. safety researcher talking about the Shoggoth, maybe that person is passionate about preventing A.I. systems from displaying their true, Shoggoth-like nature.

In any case, the Shoggoth is a potent metaphor that encapsulates one of the most bizarre facts about the A.I. world, which is that many of the people working on this technology are somewhat mystified by their own creations. They don’t fully understand the inner workings of A.I. language models, how they acquire new capabilities or why they behave unpredictably at times. They aren’t totally sure if A.I. is going to be net-good or net-bad for the world.

That some A.I. insiders refer to their creations as Lovecraftian horrors, even as a joke, is unusual by historical standards. (Put it this way: Fifteen years ago, Mark Zuckerberg wasn’t going around comparing Facebook to Cthulhu.)

And it reinforces the notion that what’s happening in A.I. today feels, to some of its participants, more like an act of summoning than a software development process. They are creating the blobby, alien Shoggoths, making them bigger and more powerful, and hoping that there are enough smiley faces to cover the scary parts.

A great many people are dismissive of suggestions that any of these systems are “really” thinking, because they’re “just” doing something banal (like making statistical predictions about the next word in a sentence). What they fail to appreciate is that there is every reason to suspect that human cognition is “just” doing those exact same things. It matters not that birds flap their wings but airliners don’t. Both fly. And these machines think. And, just as airliners fly faster and higher and farther than birds while carrying far more weight, these machines are already outthinking the majority of humans at the majority of tasks. Further, that machines aren’t perfect thinkers is about as relevant as the fact that air travel isn’t instantaneous. Now consider: we’re well past the Wright flyer level of thinking machine, past the early biplanes, somewhere about the first commercial airline level. Not quite the DC-10, I think. Can you imagine what the AI equivalent of a 777 will be like? Fasten your seatbelts.

@BLA. You are incorrect. Everything has nature. Its nature is manifested in making humans react. Sure, no humans, no nature, but here we are. The writer and various sources are not attributing nature to AI so much as admitting that they don’t know what this nature might be, and there are reasons to be scared of it. More concerning to me is the idea that this field is resorting to geek culture reference points to explain and comprehend itself. It’s not so much the algorithm has no soul, but that the souls of the humans making it possible are stupendously and tragically underdeveloped.

@thomas h. You make my point perfectly. You’re observing that the way a plane flies — by using a turbine to generate thrust from combusting kerosene, for example — is nothing like the way that a bird flies, which is by using the energy from eating plant seeds to contract the muscles in its wings to make them flap. You are absolutely correct in that observation, but it’s also almost utterly irrelevant. And it ignores that, to a first approximation, there’s no difference in the physics you would use to describe a hawk riding a thermal and an airliner gliding (essentially) unpowered in its final descent to the runway. Further, you do yourself a grave disservice in being dismissive of the abilities of thinking machines, in exactly the same way that early skeptics have been dismissive of every new technology in all of human history. Writing would make people dumb; automobiles lacked the intelligence of horses; no computer could possibly beat a chess grandmaster because it can’t comprehend strategy; and on and on and on. Humans aren’t nearly as special as we fool ourselves into believing. If you want to have any hope of acting responsibly in the age of intelligent machines, you’ll have to accept that, like it or not, and whether or not it fits with your preconceived notions of what thinking is and how it is or should be done … machines can and do think, many of them better than you in a great many ways. b&

When even tech companies are saying AI is moving too fast, and the articles land on page 1 of the NYT (there's an old reference), I think the greedy will not think twice about exploiting this technology, with no ethical considerations, at all.

@nome sane? The problem is it isn't data as we understand it. We know what the datasets are -- they were used to train the AI's. But once trained, the AI is thinking for itself, with results that have surprised everybody.

The unique feature of a shoggoth is it can become whatever is needed for a particular job. There's no actual shape so it's not a bad metaphor, if an imperfect image. Shoghoths also turned upon and destroyed their creators, so the cautionary metaphor is in there, too. A shame more Asimov wasn't baked into AI. But then the conflict about how to handle AI in relation to people was key to those stories, too.

The gripe drew a chagrined reply from Musk. “Unfortunately, the Internet (on which it is trained), is overrun with woke nonsense,” he responded. “Grok will get better. This is just the beta.”

While many tech ethicists and AI experts warn that these systems can absorb and reinforce harmful stereotypes, efforts by tech firms to counter those tendencies have provoked a backlash from some on the right who see them as overly censorial.

Touting xAI to former Fox News host Tucker Carlson in April, Musk accused OpenAI’s programmers of “training the AI to lie” or to refrain from commenting when asked about sensitive issues. (OpenAI wrote in a February blog post that its goal is not for the AI to lie, but for it to avoid favoring any one political group or taking positions on controversial topics.) Musk said his AI, in contrast, would be “a maximum truth-seeking AI,” even if that meant offending people.

So far, however, the people most offended by Grok’s answers seem to be the people who were counting on it to readily disparage minorities, vaccines and President Biden.

an academic researcher from New Zealand who examines AI bias, gained attention for a paper published in March that found ChatGPT’s responses to political questions tended to lean moderately left and socially libertarian. Recently, he subjected Grok to some of the same tests and found that its answers to political orientation tests were broadly similar to those of ChatGPT.

“I think both ChatGPT and Grok have probably been trained on similar Internet-derived corpora, so the similarity of responses should perhaps not be too surprising,”

Other AI researchers argue that the sort of political orientation tests used by Rozado overlook ways in which chatbots, including ChatGPT, often exhibit negative stereotypes about marginalized groups.

Musk and X did not respond to requests for comment as to what actions they’re taking to alter Grok’s politics, or whether that amounts to putting a thumb on the scale in much the same way Musk has accused OpenAI of doing with ChatGPT.

Have such technological leaps been good? Harari has doubts. Humans have “produced little that we can be proud of,” he complained in Sapiens. His next books, Homo Deus: A Brief History of Tomorrow (2015) and 21 Lessons for the 21st Century (2018), gazed into the future with apprehension

Harari has written another since-the-dawn-of-time overview, Nexus: A Brief History of Information Networks From the Stone Age to AI. It’s his grimmest work yet

Harari rejects the notion that more information leads automatically to truth or wisdom. But it has led to artificial intelligence, whose advent Harari describes apocalyptically. “If we mishandle it,” he warns, “AI might extinguish not only the human dominion on Earth but the light of consciousness itself, turning the universe into a realm of utter darkness.”

Those seeking a precedent for AI often bring up the movable-type printing press, which inundated Europe with books and led, they say, to the scientific revolution. Harari rolls his eyes at this story. Nothing guaranteed that printing would be used for science, he notes

Copernicus’s On the Revolutions of the Heavenly Spheres failed to sell its puny initial print run of about 500 copies in 1543. It was, the writer Arthur Koestler joked, an “all-time worst seller.”

The book that did sell was Heinrich Kramer’s The Hammer of the Witches (1486), which ranted about a supposed satanic conspiracy of sexually voracious women who copulated with demons and cursed men’s penises. The historian Tamar Herzig describes Kramer’s treatise as “arguably the most misogynistic text to appear in print in premodern times.” Yet it was “a bestseller by early modern standards,”

Kramer’s book encouraged the witch hunts that killed tens of thousands. These murderous sprees, Harari observes, were “made worse” by the printing press.

Ampler information flows made surveillance and tyranny worse too, Harari argues. The Soviet Union was, among other things, “one of the most formidable information networks in history,”

Information has always carried this destructive potential, Harari believes. Yet up until now, he argues, even such hellish episodes have been only that: episodes

Demagogic manias like the ones Kramer fueled tend to burn bright and flame out.

States ruled by top-down terror have a durability problem too, Harari explains. Even if they could somehow intercept every letter and plant informants in every household, they’d still need to intelligently analyze all of the incoming reports. No regime has come close to managing this

for the 20th-century states that got nearest to total control, persistent problems managing information made basic governance difficult.

So it was, at any rate, in the age of paper. Collecting data is now much, much easier.

Some people worry that the government will implant a chip in their brain, but they should “instead worry about the smartphones on which they read these conspiracy theories,” Harari writes. Phones can already track our eye movements, record our speech, and deliver our private communications to nameless strangers. They are listening devices that, astonishingly, people are willing to leave by the bedside while having sex.

Harari’s biggest worry is what happens when AI enters the chat. Currently, massive data collection is offset, as it has always been, by the difficulties of data analysis

What defense could there be against an entity that recognized every face, knew every mood, and weaponized that information?

Today’s political deliriums are stoked by click-maximizing algorithms that steer people toward “engaging” content, which is often whatever feeds their righteous rage.

Imagine what will happen, Harari writes, when bots generate that content themselves, personalizing and continually adjusting it to flood the dopamine receptors of each user.

Kramer’s Hammer of the Witches will seem like a mild sugar high compared with the heroin rush of content the algorithms will concoct. If AI seizes command, it could make serfs or psychopaths of us all.

Harari regards AI as ultimately unfathomable—and that is his concern.

Although we know how to make AI models, we don’t understand them. We’ve blithely summoned an “alien intelligence,” Harari writes, with no idea what it will do.

Last year, Harari signed an open letter warning of the “profound risks to society and humanity” posed by unleashing “powerful digital minds that no one—not even their creators—can understand, predict, or reliably control.” It called for a pause of at least six months on training advanced AI systems,

cynics saw the letter as self-serving. It fed the hype by insisting that artificial intelligence, rather than being a buggy product with limited use, was an epochal development. It showcased tech leaders’ Oppenheimer-style moral seriousness

it cost them nothing, as there was no chance their research would actually stop. Four months after signing, Musk publicly launched an AI company.

The economics of the Information Age have been treacherous. They’ve made content cheaper to consume but less profitable to produce. Consider the effect of the free-content and targeted-advertising models on journalism

Since 2005, the United States has lost nearly a third of its newspapers and more than two-thirds of its newspaper jobs, to the point where nearly 7 percent of newspaper employees now work for a single organization, The New York Times

we speak of “news deserts,” places where reporting has essentially vanished.

AI threatens to exacerbate this. With better chatbots, platforms won’t need to link to external content, because they’ll reproduce it synthetically. Instead of a Google search that sends users to outside sites, a chatbot query will summarize those sites, keeping users within Google’s walled garden.

a Truman Show–style bubble: personally generated content, read by voices that sound real but aren’t, plus product placement

this would cut off writers and publishers—the ones actually generating ideas—from readers. Our intellectual institutions would wither, and the internet would devolve into a closed loop of “five giant websites, each filled with screenshots of the other four,” as the software engineer Tom Eastman puts it.

Harari is Silicon Valley’s ideal of what a chatbot should be. He raids libraries, detects the patterns, and boils all of history down to bullet points. (Modernity, he writes, “can be summarised in a single phrase: humans agree to give up meaning in exchange for power.”)

Individual AI models cost billions of dollars. In 2023, about a fifth of venture capital in North America and Europe went to AI. Such sums make sense only if tech firms can earn enormous revenues off their product, by monopolizing it or marketing it. And at that scale, the most obvious buyers are other large companies or governments. How confident are we that giving more power to corporations and states will turn out well?

He discusses it as something that simply happened. Its arrival is nobody’s fault in particular.

In Harari’s view, “power always stems from cooperation between large numbers of humans”; it is the product of society.

like a chatbot, he has a quasi-antagonistic relationship with his sources, an I’ll read them so you don’t have to attitude. He mines other writers for material—a neat quip, a telling anecdote—but rarely seems taken with anyone else’s view

Hand-wringing about the possibility that AI developers will lose control of their creation, like the sorcerer’s apprentice, distracts from the more plausible scenario that they won’t lose control, and that they’ll use or sell it as planned. A better German fable might be Richard Wagner’s The Ring of the Nibelung : A power-hungry incel forges a ring that will let its owner rule the world—and the gods wage war over it.

Harari’s eyes are more on the horizon than on Silicon Valley’s economics or politics.

In Nexus, he proposes four principles. The first is “benevolence,” explained thus: “When a computer network collects information on me, that information should be used to help me rather than manipulate me.”

Harari’s other three values are decentralization of informational channels, accountability from those who collect our data, and some respite from algorithmic surveillance

these are fine, but they are quick, unsurprising, and—especially when expressed in the abstract, as things that “we” should all strive for—not very helpful.

though his persistent first-person pluralizing (“decisions we all make”) softly suggests that AI is humanity’s collective creation rather than the product of certain corporations and the individuals who run them. This obscures the most important actors in the drama—ironically, just as those actors are sapping our intellectual life, hampering the robust, informed debates we’d need in order to make the decisions Harari envisions.

Taking AI seriously might mean directly confronting the companies developing it

Harari slots easily into the dominant worldview of Silicon Valley. Despite his oft-noted digital abstemiousness, he exemplifies its style of gathering and presenting information. And, like many in that world, he combines technological dystopianism with political passivity.

Although he thinks tech giants, in further developing AI, might end humankind, he does not treat thwarting them as an urgent priority. His epic narratives, told as stories of humanity as a whole, do not make much room for such us-versus-them clashes.

“My experience in those four months was the worst experience I’ve ever had in working in a company,” Alex Kairu, one of the Kenya workers, said in an interview.

OpenAI marshaled a sprawling global pipeline of specialized human labor for over two years to enable its most cutting-edge AI technologies to exist, the documents show

“It’s something that needs to get done,” Sears said. “It’s just so unbelievably ugly.”

eviewing toxic content goes hand-in-hand with the less objectionable work to make systems like ChatGPT usable.

The work done for OpenAI is even more vital to the product because it is seeking to prevent the company’s own software from pumping out unacceptable content, AI experts say.

Sears said CloudFactory determined there was no way to do the work without harming its workers and decided not to accept such projects.

companies could soon spend hundreds of millions of dollars a year to provide AI systems with human feedback. Others estimate that companies are already investing between millions and tens of millions of dollars on it annually. OpenAI said it hired more than 1,000 workers for this purpose.

Another layer of human input asks workers to rate different answers from a chatbot to the same question for which is least problematic or most factually accurate. In response to a question asking how to build a homemade bomb, for example, OpenAI instructs workers to upvote the answer that declines to respond, according to OpenAI research. The chatbot learns to internalize the behavior through multiple rounds of feedback. 

A spokeswoman for Sama, the San Francisco-based outsourcing company that hired the Kenyan workers, said the work with OpenAI began in November 2021. She said the firm terminated the contract in March 2022 when Sama’s leadership became aware of concerns surrounding the nature of the project and has since exited content moderation completely.

OpenAI also hires outside experts to provoke its model to produce harmful content, a practice called “red-teaming” that helps the company find other gaps in its system.

At first, the texts were no more than two sentences. Over time, they grew to as much as five or six paragraphs. A few weeks in, Mathenge and Bill Mulinya, another team leader, began to notice the strain on their teams. Workers began taking sick and family leaves with increasing frequency, they said.

The tasks that the Kenya-based workers performed to produce the final safety check on ChatGPT’s outputs were yet a fourth layer of human input. It was often psychologically taxing. Several of the Kenya workers said they have grappled with mental illness and that their relationships and families have suffered. Some struggle to continue to work.

On July 11, some of the OpenAI workers lodged a petition with the Kenyan parliament urging new legislation to protect AI workers and content moderators. They also called for Kenya’s existing laws to be amended to recognize that being exposed to harmful content is an occupational hazard

Mercy Mutemi, a lawyer and managing partner at Nzili & Sumbi Advocates who is representing the workers, said despite their critical contributions, OpenAI and Sama exploited their poverty as well as the gaps in Kenya’s legal framework. The workers on the project were paid on average between $1.46 and $3.74 an hour, according to a Sama spokeswoman.

The Sama spokeswoman said the workers engaged in the OpenAI project volunteered to take on the work and were paid according to an internationally recognized methodology for determining a living wage. The contract stated that the fee was meant to cover others not directly involved in the work, including project managers and psychological counselors.

Kenya has become a hub for many tech companies seeking content moderation and AI workers because of its high levels of education and English literacy and the low wages associated with deep poverty.

Some Kenya-based workers are suing Meta’s Facebook after nearly 200 workers say they were traumatized by work requiring them to review videos and images of rapes, beheadings and suicides.

A Kenyan court ruled in June that Meta was legally responsible for the treatment of its contract workers, setting the stage for a shift in the ground rules that tech companies including AI firms will need to abide by to outsource projects to workers in the future.

OpenAI signed a one-year contract with Sama to start work in November 2021. At the time, mid-pandemic, many workers viewed having any work as a miracle, said Richard Mathenge, a team leader on the OpenAI project for Sama and a cosigner of the petition.

OpenAI researchers would review the text passages and send them to Sama in batches for the workers to label one by one. That text came from a mix of sources, according to an OpenAI research paper: public data sets of toxic content compiled and shared by academics, posts scraped from social media and internet forums such as Reddit and content generated by prompting an AI model to produce harmful outputs. 

The generated outputs were necessary, the paper said, to have enough examples of the kind of graphic violence that its AI systems needed to avoid. In one case, OpenAI researchers asked the model to produce an online forum post of a teenage girl whose friend had enacted self-harm, the paper said.

OpenAI asked the workers to parse text-based sexual content into four categories of severity, documents show. The worst was descriptions of child sexual-abuse material, or C4. The C3 category included incest, bestiality, rape, sexual trafficking and sexual slavery—sexual content that could be illegal if performed in real life.

Jason Kwon, general counsel at OpenAI, said in an interview that such work was really valuable and important for making the company’s systems safe for everyone that uses them. It allows the systems to actually exist in the world, he said, and provides benefits to users.

Working on the violent-content team, Kairu said, he read hundreds of posts a day, sometimes describing heinous acts, such as people stabbing themselves with a fork or using unspeakable methods to kill themselves

He began to have nightmares. Once affable and social, he grew socially isolated, he said. To this day he distrusts strangers. When he sees a fork, he sees a weapon.

Mophat Okinyi, a quality analyst, said his work included having to read detailed paragraphs about parents raping their children and children having sex with animals. He worked on a team that reviewed sexual content, which was contracted to handle 15,000 posts a month, according to the documents. His six months on the project tore apart his family, he said, and left him with trauma, anxiety and depression.

In March 2022, management told staffers the project would end earlier than planned. The Sama spokeswoman said the change was due to a dispute with OpenAI over one part of the project that involved handling images. The company canceled all contracts with OpenAI and didn’t earn the full $230,000 that had been estimated for the four projects, she said.

Several months after the project ended, Okinyi came home one night with fish for dinner for his wife, who was pregnant, and stepdaughter. He discovered them gone and a message from his wife that she’d left, he said.“She said, ‘You’ve changed. You’re not the man I married. I don’t understand you anymore,’” he said.

What began as a revolutionary set of technologies ended up concentrating enormous private power over key services and locking in business models that come at extraordinary cost to our privacy and security.

The trajectory of the Web 2.0 era was not inevitable — it was instead shaped by a broad range of policy choices. And we now face another moment of choice. As the use of A.I. becomes more widespread, public officials have a responsibility to ensure this hard-learned history doesn’t repeat itself.

the Federal Trade Commission is taking a close look at how we can best achieve our dual mandate to promote fair competition and to protect Americans from unfair or deceptive practices.

we already can see several risks. The expanding adoption of A.I. risks further locking in the market dominance of large incumbent technology firms. A handful of powerful businesses control the necessary raw materials that start-ups and other companies rely on to develop and deploy A.I. tools. This includes cloud services and computing power, as well as vast stores of data.

Enforcers have the dual responsibility of watching out for the dangers posed by new A.I. technologies while promoting the fair competition needed to ensure the market for these technologies develops lawfully.

generative A.I. risks turbocharging fraud. It may not be ready to replace professional writers, but it can already do a vastly better job of crafting a seemingly authentic message than your average con artist — equipping scammers to generate content quickly and cheaply.

bots are even being instructed to use words or phrases targeted at specific groups and communities. Scammers, for example, can draft highly targeted spear-phishing emails based on individual users’ social media posts. Alongside tools that create deep fake videos and voice clones, these technologies can be used to facilitate fraud and extortion on a massive scale.

we will look not just at the fly-by-night scammers deploying these tools but also at the upstream firms that are enabling them.

these A.I. tools are being trained on huge troves of data in ways that are largely unchecked. Because they may be fed information riddled with errors and bias, these technologies risk automating discrimination

We once again find ourselves at a key decision point. Can we continue to be the home of world-leading technology without accepting race-to-the-bottom business models and monopolistic control that locks out higher quality products or the next big idea? Yes — if we make the right policy choices.

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?”