AI 2027

Nuclear deterrence -- human cloning -- bioweapon proliferation -- Antarctic neutrality -- the list goes on.

> How successful do you think that political organization is going to be in slowing China's progress?

I wish people would stop with this tired war-mongering. China was not the one who opened up this can of worms. China has never been the one pushing the edge of capabilities. Before Sam Altman decided to give ChatGPT to the world, they were actively cracking down on software companies (in favor of hardware & "concrete" production).

We, the US, are the ones who chose to do this. We started the race. We put the world, all of humanity, on this path.

> Do you honestly think that would make a lick of difference?

I don't know, it depends. Perhaps we're lucky and the timelines are slow enough that 20-30% of the population loses their jobs before things become unrecoverable. Tech companies used to warn people not to wear their badges in public in San Francisco -- and that was what, 2020? Would you really want to work at "Human Replacer, Inc." when that means walking out and about among a population who you know hates you, viscerally? Or if we make it to 2028 in the same condition. The Bonus Army was bad enough -- how confident are you that the government would stand their ground, keep letting these labs advance capabilities, when their electoral necks were on the line?

This defeatism is a self-fulfilling prophecy. The people have the power to make things happen, and rhetoric like this is the most powerful thing holding them back.

https://www.theguardian.com/technology/2017/apr/18/god-in-th...

In section 4 they discuss their projections specifically for model size, the state of inference chips in 2027, etc. It's largely pretty in line with expectations in terms of the capacity, and they only project them using 10k of their latest gen wafer scale inference chips by late 2027, roughly like 1M H100 equivalents. That doesn't seem at all impossible. They also earlier on discuss expectations for growth in efficiency of chips, and for growth in spending, which is only ~10x over the next 2.5 years, not unreasonable in absolute terms at all given the many tens of billions of dollars flooding in.

So on the "can we train the AI" front, they mostly are just projecting 2.5 years of the growth in scale we've been seeing.

The reason they predict a fairly hard takeoff is they expect that distillation, some algorithmic improvements, and iterated creation of synthetic data, training, and then making more synthetic data will enable significant improvements in efficiency of the underlying models (something still largely in line with developments over the last 2 years). In particular they expect a 10T parameter model in early 2027 to be basically human equivalent, and they expect it to "think" at about the rate humans do, 10 words/second. That would require ~300 teraflops of compute per second to think at that rate, or ~0.1H100e. That means one of their inference chips could potentially run ~1000 copies (or fewer copies faster etc. etc.) and thus they have the capacity for millions of human equivalent researchers (or 100k 40x speed researchers) in early 2027.

They further expect distillation of such models etc. to squeeze the necessary size down / more expensive models overseeing much smaller but still good models squeezing the effective amount of compute necessary, down to just 2T parameters and ~60 teraflops each, or 5000 human-equivalents per inference chip, making for up to 50M human-equivalents by late 2027.

This is probably the biggest open question and the place where the most criticism seems to me to be warranted. Their hardware timelines are pretty reasonable, but one could easily expect needing 10-100x more compute or even perhaps 1000x than they describe to achieve Nobel-winner AGI or superintelligence.