Former OpenAI safety researcher brands pace of AI development ‘terrifying’

Tue, 28 Jan 2025, 19:06
Full Content
Full article content available

A former safety researcher at OpenAI said he is “pretty terrified” about the pace of development in artificial intelligence, warning the industry is taking a “very risky gamble” on the technology.

Steven Adler expressed concerns about companies seeking to rapidly develop artificial general intelligence (AGI), a theoretical term referring to systems that match or exceed humans at any intellectual task.

Adler, who left OpenAI in November, said in a series of posts on X that

he’d had a “wild ride”

at the US company and will miss “many parts of it”.

However, he said the technology was developing so quickly it

raised doubts

about the

future of humanity

.

“I’m pretty terrified by the pace of AI development these days,” he said. “When I think about where I’ll raise a future family, or how much to save for retirement, I can’t help but wonder: will humanity even make it to that point?”

Some experts,

such as Nobel prize winner Geoffrey Hinton

, fear that powerful AI systems could evade human control with potentially catastrophic consequences. Others, such as Meta’s chief AI scientist, Yann LeCun, have

played down the existential threat

, saying AI “could actually save humanity from extinction”.

According to Adler’s

LinkedIn profile

, he led safety-related research for “first-time product launches” and “more speculative long-term AI systems” in a four-year career at OpenAI.

Referring to the development of AGI, OpenAI’s core goal, Adler added: “An AGI race is a very risky gamble, with huge downside.” Adler said no research lab had a solution to AI alignment – the process of ensuring that systems adhere to a set of human values – and that the industry might be moving too fast to find one.

“The faster we race, the less likely that anyone finds one in time.”

Adler’s X posts came as China’s DeepSeek, which is also seeking to develop AGI,

rattled the US tech industry

by unveiling a model that rivalled OpenAI’s technology despite being developed with apparently fewer resources.

Warning that the industry appeared to be “stuck in a really bad equilibrium”, Adler said “real safety regs” were needed.

“Even if a lab truly wants to develop AGI responsibly, others can still cut corners to catch up, maybe disastrously.”

Adler and OpenAI have been contacted for comment.

AI Model Selection

Avg. Response: 10.0s

Llama3.2:1b

Meta
Default
Size: 1B
Success Rate: 100.0%

Llama-3.2-1B-Instruct-GGUF

Meta
Size: 1B
Success Rate: 100.0%
All models run locally on our servers. Response times may vary based on server load.