The 2 nations might doubtlessly attain slender agreements the place their pursuits converge – akin to prohibiting the event of AI-enabled bioweapons – however a broader consensus on slowing the tempo of AI improvement is much less seemingly, Prof Hanh mentioned.
“A broad settlement runs into an actual zero-sum drawback. In case you maintain again and the opposite aspect does not, you lose,” he informed CNA.
Mr Enriquez mentioned diplomats might even have confronted a neater process negotiating Chilly Warfare-era nuclear treaties than they’d making an attempt to incentivise a world AI slowdown.
Not like the nuclear arms race, the AI race is just not pushed solely by army dominance. Financial acquire can be a significant incentive, with proponents arguing that AI will speed up financial development, create jobs and generate worth.
Whereas the Chilly Warfare nuclear race revolved round deterrence and the idea of mutually assured destruction, the worth proposition of the AI race is completely different, Mr Enriquez mentioned.
The vary of events driving the competitors can be far broader.
“This isn’t solely statewide competitors. That is additionally company {industry} competitors, and that may get actually ruthless as nicely,” he mentioned.
“The record of gamers that must be aligned is a lot larger than what they had been when the nuclear treaties had been being signed.”
Mr Ang mentioned the present degree of mistrust between nations might imply cooperation emerges solely after a extra severe incident.
“It might take a significant AI-related incident, with severe influence to life, to deliver them collectively to cooperate, assuming it doesn’t drive them additional aside,” he mentioned.
Are main AI firms exaggerating the dangers to gradual competitors?
There may be additionally a extra cynical interpretation of the requires warning.
Critics have questioned whether or not the current warnings from Anthropic and OpenAI are partly meant to generate hype across the capabilities of their merchandise.
Each firms had been anticipated to launch their first IPOs this 12 months, though OpenAI mentioned final week that it could not go public in 2026 due to security considerations surrounding AI.
An industry-wide slowdown might additionally shield at this time’s main corporations from new competitors, Mr Ang identified.
That doesn’t imply efforts to gradual improvement needs to be rejected, he mentioned. However any internationally agreed guidelines and enforcement mechanisms would want to make sure that the most important gamers themselves stay topic to scrutiny.
“Whereas techno-libertarians usually argue that guidelines gradual innovation, and techno-determinists usually argue that know-how can’t be managed, the truth is that society has at all times needed to develop guidelines, legislation, and norms to form how we use know-how,” he mentioned.
Mr Enriquez mentioned that even when a few of the dangers raised over the previous week show to be overstated, stronger security infrastructure and unbiased oversight of AI improvement are nonetheless wanted.
“We’re in a second the place not less than they (main AI firms) appear like they’re open to that regulation … (and) oversight, which is sorely missing,” he mentioned.
“It might be remiss to not capitalise on the second.”
On the identical time, specialists mentioned it was vital to contemplate who would finally profit from a slowdown.
Bigger firms have the cash, authorized assets and affect to soak up compliance prices extra simply than smaller gamers, mentioned Mr Dutt.
“If the end result (of a slowdown) is a rulebook smaller gamers can’t meet, the language of security begins functioning as a market barrier,” he mentioned.
“That doesn’t make each name for warning cynical, however it’s price asking every time: who is definitely being requested to decelerate, and who continues to be free to maintain shifting?”
