GPT/LLM links, 8/22

Ian Bremmer and Mustafa Suleyman on world power relationships; Freddie deBoer is not impressed with the latest AI; Tyler Cowen on limits to growth; The Zvi disagrees with Tyler

Ian Bremmer and Mustafa Suleyman write,

Soon, AI developers will likely succeed in creating systems with self-improving capabilities—a critical juncture in the trajectory of this technology that should give everyone pause.

This assumption is at the core of doomer scenarios.

There is little use in regulating AI in some countries if it remains unregulated in others. Because AI can proliferate so easily, its governance can have no gaps.

To me, this sounds like they are saying that regulation of AI must be totalitarian in order to succeed. That in itself is a doomer scenario. But

at least for the next few years, AI’s trajectory will be largely determined by the decisions of a handful of private businesses, regardless of what policymakers in Brussels or Washington do. In other words, technologists, not policymakers or bureaucrats, will exercise authority over a force that could profoundly alter both the power of nation-states and how they relate to each other.

The authors wish to conjure up regulatory bodies that are worldwide, wise, and adaptable. Predictably, Robert Wright likes the notion. If they are correct that AI necessitates this manner of regulation, then count me as a doomer.

Freddie deBoer writes,

it feels like people have been told so relentlessly by the media that what we are choosing to call artificial intelligence is currently, right now, already amazing that they feel compelled to go along with it. But this isn’t amazing. It’s a demonstration of the profound limitations of these systems that people are choosing to see as a representation of their strengths.

He makes it sound like a naked emperor.

Tyler Cowen writes,

as one set of constraints is relaxed — in this case access to intelligence — the remaining constraints will matter all the more. Regulatory delays will be more frustrating, for instance, as they will be holding back a greater amount of cognitive horsepower than in times past. Or as AI improves at finding new and better medical hypotheses to test, the lags and delays in systems for clinical trials will become all the more painful. In fact they may worsen as the system is flooded with conjectures.

And if we get the sort of world government that folks like Bremmer, Suleyman, and Robert Wright hope for, all bets are off.

Tyler expects “advanced artificial intelligence will boost the annual US growth rate by one-quarter to one-half of a percentage point.” That sounds right to me. It is a significant change, but not an earth-shattering one. The problem is that there are so many other adaptations needed to foster the economic impact of AI.

But Zvi Mowshowitz disagrees.

I think what Tyler predicts here is on the extreme low end, even if we got no further substantial foundational advances from AI beyond the GPT-4 level, and even if rather harsh restrictions are put in place. The comparisons to the Industrial Revolution continue to point if anything to far faster growth, since you would then have a metaphorical power law ordering of speed of impact from something like humans, then agriculture, then industry.

Let’s say that the industrial revolution increased the annual growth rate from 0.2 percent to 2 percent. Does a comparable increase mean going from 2 percent to 4 percent, or from 2 percent to 20 percent?

substacks referenced above:

@

AI #25: Inflection Point

@

Does AI Just Suck?