More than 1,000 artificial intelligence experts urge delay until world can be confident ‘effects will be positive and risks manageable’
More than 1,000 artificial intelligence experts, researchers and backers have joined a call for an immediate pause on the creation of “giant” AIs for at least six months, so the capabilities and dangers of systems such as GPT-4 can be properly studied and mitigated.
The demand is made in an open letter signed by major AI players including: Elon Musk, who co-founded OpenAI, the research lab responsible for ChatGPT and GPT-4; Emad Mostaque, who founded London-based Stability AI; and Steve Wozniak, the co-founder of Apple.
Its signatories also include engineers from Amazon, DeepMind,
Google, Meta and Microsoft, as well as academics including the cognitive
scientist Gary Marcus.
“Recent months have seen AI labs locked in
an out-of-control race to develop and deploy ever more powerful digital
minds that no one – not even their creators – can understand, predict, or
reliably control,” the letter says.
“Powerful AI systems should
be developed only once we are confident that their effects will be positive
and their risks will be manageable.”
The authors, coordinated by
the “longtermist” thinktank the Future of Life Institute, cite OpenAI’s own
co-founder Sam Altman in justifying their calls.
In a post from
February, Altman wrote: “At some point, it may be important to get
independent review before starting to train future systems, and for the most
advanced efforts to agree to limit the rate of growth of compute used for
creating new models.”
The letter continued: “We agree. That point is now.”
If
researchers will not voluntarily pause their work on AI models more powerful
than GPT-4, the letter’s benchmark for “giant” models, then “governments
should step in”, the authors say.
“This does not mean a pause on
AI development in general, merely a stepping back from the dangerous race to
ever-larger unpredictable black-box models with emergent capabilities,” they
add.
Since the release of GPT-4, OpenAI has been adding
capabilities to the AI system with “plugins”, giving it the ability to look
up data on the open web, plan holidays, and even order groceries. But the
company has to deal with “capability overhang”: the issue that its own
systems are more powerful than it knows at release.
As researchers experiment with GPT-4 over the coming weeks and
months, they are likely to uncover new ways of “prompting” the system that
improve its ability to solve difficult problems.
One recent
discovery was that the AI is noticeably more accurate at answering questions
if it is first told to do so “in the style of a knowledgable expert”.
The
call for strict regulation stands in stark contrast to the UK government’s
flagship AI regulation white paper, published on Wednesday, which contains
no new powers at all.
Instead, the government says, the focus is
on coordinating existing regulators such as the Competition and Markets
Authority and Health and Safety Executive, offering five “principles”
through which they should think about AI.
“Our new approach is based on strong principles so that people
can trust businesses to unleash this technology of tomorrow,” said the
science, innovation and technology secretary, Michelle Donelan.
The
Ada Lovelace Institute was among those that criticised the announcement.
“The UK’s approach has significant gaps, which could leave harms
unaddressed, and is underpowered relative to the urgency and scale of the
challenge,” said Michael Birtwistle, who leads data and AI law and policy at
the research institute.
“The government’s timeline of a year or
more for implementation will leave risks unaddressed just as AI systems are
being integrated at pace into our daily lives, from search engines to office
suite software.”
Labour joined the criticism, with the shadow
culture secretary, Lucy Powell, accusing the government of “letting down
their side of the bargain”.
She said: “This regulation will take months, if not years, to
come into effect. Meanwhile, ChatGPT, Google’s Bard and many others are
making AI a regular part of our everyday lives.
“The government
risks re-enforcing gaps in our existing regulatory system, and making the
system hugely complex for businesses and citizens to navigate, at the same
time as they’re weakening those foundations through their upcoming data
bill.”