Group items matching

in title, tags, annotations or url

They also claim that OpenAI has used hardball tactics to prevent workers from voicing their concerns about the technology, including restrictive nondisparagement agreements that departing employees were asked to sign.

“OpenAI is really excited about building A.G.I., and they are recklessly racing to be the first there,” said Daniel Kokotajlo, a former researcher in OpenAI’s governance division and one of the group’s organizers.

Other members include William Saunders, a research engineer who left OpenAI in February, and three other former OpenAI employees: Carroll Wainwright, Jacob Hilton and Daniel Ziegler. Several current OpenAI employees endorsed the letter anonymously because they feared retaliation from the company,

At OpenAI, Mr. Kokotajlo saw that even though the company had safety protocols in place — including a joint effort with Microsoft known as the “deployment safety board,” which was supposed to review new models for major risks before they were publicly released — they rarely seemed to slow anything down.

So was the departure of Dr. Leike, who along with Dr. Sutskever had led OpenAI’s “superalignment” team, which focused on managing the risks of powerful A.I. models. In a series of public posts announcing his departure, Dr. Leike said he believed that “safety culture and processes have taken a back seat to shiny products.”

“When I signed up for OpenAI, I did not sign up for this attitude of ‘Let’s put things out into the world and see what happens and fix them afterward,’” Mr. Saunders said.

Mr. Kokotajlo, 31, joined OpenAI in 2022 as a governance researcher and was asked to forecast A.I. progress. He was not, to put it mildly, optimistic.In his previous job at an A.I. safety organization, he predicted that A.G.I. might arrive in 2050. But after seeing how quickly A.I. was improving, he shortened his timelines. Now he believes there is a 50 percent chance that A.G.I. will arrive by 2027 — in just three years.

He also believes that the probability that advanced A.I. will destroy or catastrophically harm humanity — a grim statistic often shortened to “p(doom)” in A.I. circles — is 70 percent.

Last month, two senior A.I. researchers — Ilya Sutskever and Jan Leike — left OpenAI under a cloud. Dr. Sutskever, who had been on OpenAI’s board and voted to fire Mr. Altman, had raised alarms about the potential risks of powerful A.I. systems. His departure was seen by some safety-minded employees as a setback.

Mr. Kokotajlo said, he became so worried that, last year, he told Mr. Altman that the company should “pivot to safety” and spend more time and resources guarding against A.I.’s risks rather than charging ahead to improve its models. He said that Mr. Altman had claimed to agree with him, but that nothing much changed.

In April, he quit. In an email to his team, he said he was leaving because he had “lost confidence that OpenAI will behave responsibly" as its systems approach human-level intelligence.

“The world isn’t ready, and we aren’t ready,” Mr. Kokotajlo wrote. “And I’m concerned we are rushing forward regardless and rationalizing our actions.”

On his way out, Mr. Kokotajlo refused to sign OpenAI’s standard paperwork for departing employees, which included a strict nondisparagement clause barring them from saying negative things about the company, or else risk having their vested equity taken away.

Many employees could lose out on millions of dollars if they refused to sign. Mr. Kokotajlo’s vested equity was worth roughly $1.7 million, he said, which amounted to the vast majority of his net worth, and he was prepared to forfeit all of it.

Mr. Altman said he was “genuinely embarrassed” not to have known about the agreements, and the company said it would remove nondisparagement clauses from its standard paperwork and release former employees from their agreements.)

In their open letter, Mr. Kokotajlo and the other former OpenAI employees call for an end to using nondisparagement and nondisclosure agreements at OpenAI and other A.I. companies.

“Broad confidentiality agreements block us from voicing our concerns, except to the very companies that may be failing to address these issues,”

They also call for A.I. companies to “support a culture of open criticism” and establish a reporting process for employees to anonymously raise safety-related concerns.

They have retained a pro bono lawyer, Lawrence Lessig, the prominent legal scholar and activist

Mr. Kokotajlo and his group are skeptical that self-regulation alone will be enough to prepare for a world with more powerful A.I. systems. So they are calling for lawmakers to regulate the industry, too.

“There needs to be some sort of democratically accountable, transparent governance structure in charge of this process," Mr. Kokotajlo said. “Instead of just a couple of different private companies racing with each other, and keeping it all secret.”

Analyzing data covering a near universe of Americans born from 1978 to 1992, the researchers found that when employment among the poor parents of children in a community improves, those children are better off economically as adults

it doesn’t rely on whether a child’s own parents are employed: Outcomes also improve for children who simply grow up in a neighborhood where more parents have jobs. In other words, their own parents might be unemployed, but if their schoolmates’ parents work, their outcomes will be better.

: In places where parental employment deteriorates, the opposite happens—children do worse as adults.

“Growing up in a community where employment rates are higher for people in your race and class—if those employment rates are higher, the kids who grew up in those environments do better in the long run,”

For poor white children born in 1978, vast swaths of the U.S. were a land of opportunity. Apart from some areas, such as Appalachia and Rust Belt areas of Michigan and Ohio, these children overall had a good chance of making it to a higher rung on the income ladder than their parents. 

For poor white children born in 1992, the map was more constrained. While some parts of America, such as the upper Midwest, remained bastions of opportunity, much of the U.S. did worse. Children with parents at the 25th percentile who grew up in Milwaukee had lower income when they turned 27 in 2019—an inflation-adjusted $30,619—than their older counterparts had.

But poor Black children born in 1992 who grew up in Kent County did better. On average, a Black child at the 25th percentile in that birth cohort had inflation-adjusted household income of $23,547 at age 27, or 38% more than their 1978 counterpart. 

In contrast to the environment for poor white children, few places in America provided poor Black children born in 1978 much opportunity for advancement.

Kent County, Mich., is a little more than 100 miles directly east of Milwaukee, and home to Grand Rapids, another old-line manufacturing city. Black children born to parents at the 25th percentile by income in 1978 who grew up in Kent had an average household income of $17,029 at age 27. That put them at the 28th percentile by income, leaving them on essentially the same rung as their parents.

This dynamic, playing out across the country, led to a significant widening of the income gap between poor and well-off white children. A white child born to parents at the 25th percentile in 1978 made, on average, an inflation-adjusted $10,383 less at age 27 than a child born to parents at the 75th percentile. But for children born in 1992, that income difference was 27% larger at $13,202.

One thing that changed for poor white children compared with poor Black children was the relative share of their parents who were employed. 

In Milwaukee County, for example, the share of the parents of low-income white children who were employed went from 69.1% for children born in 1978 to 55.6% for children born in 1992—a 13.5 percentage point decline. Those employment declines were likely driven by the loss of more than a third of the county’s manufacturing jobs during those years. Children born at the 25th percentile in 1992 reached the 44th percentile on average, 4 percentage points lower than those born in 1978.

And this pattern of falling parental employment and worsening outcomes repeated itself across much of the country. In counties where employment rates among poor white parents fell sharply, such as Philadelphia County, Pa., the income ranks of poor white children in adulthood fell too. Mobility tended not to deteriorate as much for poor white children growing up in counties such as Sumner, Tenn., where parental employment held up better.

the experience in Kent was the general pattern across the country, with employment rates for the parents of poor Black children falling less than their poor white counterparts, or gaining. And as was the case with poor whites, the new research finds that changes in parental employment rates in a community were strongly associated with changes in children’s outcomes.

the researchers emphasize that it is not a zero-sum game, with mobility for poor Black people improving only because mobility for poor white people is falling: Indeed, places where opportunity for Black children improved most are also where white children did best.