AI Global Odyssey: REAL Situational Awareness - #63

Steve talks about AI in light of his recent travels to SF, Singapore, Manila, Berkeley, and Silicon Valley.
Chapters:
0:00 Overview: SF, Singapore, Manila, Berkeley
02:20 AI Bubble in Silicon Valley
04:22 Scaling Laws and AGI
24:14 Global AI: Singapore, Philippines, real Enterprise applications
35:37 AGI: Manhattan Project? Manifest and P(doomers), Situational Awareness
52:03 China LLMs, Huawei vs Nvidia GPUs, US vs China AI race
Links:
Scaling Laws for Neural Language Models: arxiv.org/pdf/2001.08361
AI rollout in Philippines Call Centers: • Preparing BPOs for the...
LLM rankings and Qwen2: huggingface.co/spaces/open-ll...
The Economist on China LLMs: archive.ph/nW7ch
Situational Awareness summary: x.com/hsu_steve/status/180341...
Music used with permission from Blade Runner Blues Livestream improvisation by State Azure.
-
Steve Hsu is Professor of Theoretical Physics and of Computational Mathematics, Science, and Engineering at Michigan State University. Previously, he was Senior Vice President for Research and Innovation at MSU and Director of the Institute of Theoretical Science at the University of Oregon. Hsu is a startup founder (SuperFocus.ai., SafeWeb, Genomic Prediction, Othram) and advisor to venture capital and other investment firms. He was educated at Caltech and Berkeley, was a Harvard Junior Fellow, and has held faculty positions at Yale, the University of Oregon, and MSU.
Please send any questions or suggestions to [e-mail] or Steve on Twitter @hsu_steve.

[@PhilipWong55]

The bubble always bursts,
What goes up must come down,
The spinning wheel must go round,
In this cycle, truth is found.

[@Michael-ei3vy]

Excellent episode.

[@Abe41194]

**Full disclosure, the following is the rambling of someone who works in tech and loves to solve hard problems. I find some of the AI stuff cool, fascinating at the least.**
33:12 This is the key point. I have yet to see anyone address this issue thoughtfully or logically for that matter. I think history shows us that the "interesting dynamic" Steve refers to is the fact that those capable of capitalizing on (here it's AI) will do so at the expense of human beings. "Different interests" to me is "making crap ton of dollas baby".
Altman et al. have said "Oh yes, well, humans will just find more meaningful work." What does that mean? What does that society look like? Spending time with their children? Building...stuff? Will they be able to afford housing/food? UBI? Has anyone presented anything remotely tangible or actionable? What will give their lives meaning after the initial honeymoon phase of not being in the rat race ends? Putting the cart before the horse and filling it with bags of $$$ will surely please the very few, not the masses.
Let's say we attempt to "reskill" entire industries, e.g. the large proportion of the people working in these call centers in the Phillipines. Assuming these models reach a threshold where they will be able to improve themselves, or a small group of human engineers will be able to, will the rate of re-skilling effectively be obsolete and swallowed by the rate of model improvement? This gets into the philosophical of course, but I fear we are potentially closer to needing an answer than further...

[@flashmo7]

The other guy talks like a "handler" . .

[@StraightAcross]

The human brain was made by god, you cannot recreate it with your machines. A folly of featless pursuits.

[@jonathanedwardgibson]

What do you do when the orbital Ai only hires the very-best robotic labor and your hungry whelps & spouse on a street corner?
Remember your blasé dismissal of local economies while global-jetting for that 10% GDP ‘someone’ gets for raiding the-commons, you know, those Talking Cows - once known as labor? Good thing money is free to helicopter in/out and not people, or there would be no escaping poverty, heh? Who wants hungry people hanging around that spacious backyard or cluttering water-features?
I appreciate the current-assessments. Confirms business logic is just as wretched and driven-mad as ever.
We are more than DNA code, which appears more vestigial than the big genome pharma bets ever warranted. Michael Levin’s work on cellular electrical networks of breed-able flatworms show genetics completely ignored by other morphology signals - long articulated by Rupert Sheldrake.

[@ai._m]

Agree, and hopefully we are very wrong about this. No one is talking about this except a few scared developers on Twitter.