alt Hacker NewsI really don't understand the national security argument. If you really do fear some fundamental breakthrough in AI from China, what's cheaper, $500 billion to rush to get there first, or spending a few billion (and likely much less) in basic research in physics, materials science, and electronics, mixed with a little bit of espionage, mixed with improving the electric grid and eliminating (or greatly reducing) fossil fuels?
Ultimately, the breakthrough in AI is going to either come from eliminating bottlenecks in computing such that we can simulate many more neurons much more cheaply (in other words, 2025-level technology scaled up is not going to really be necessary or sufficient), or some fundamental research discovery such as a new transformer paradigm. In any case, it feels like these are theoretical discoveries that, whoever makes them first, the other "side" can trivially steal or absorb the information.
Replies
no... one more lane will fix the traffic. Truly American approach
Amazing to see how DeepSeek R1 is doing better than OpenAI models with much less resources
500B of compute infrastructure with order of magnitude greater deprecation / need to return on capital. Compute isn't concrete infra with 50+ years of value, more like 5 years, i.e. need to produce 50-100B worth value per year to break even. On top of the “$125B hole that needs to be filled for each year of CapEx at today’s levels” according to Sequoia. All which with could be wiped by PRC pacing SOTA model, unless 500B of more compute will lead to qualititive model differences. Right now, I don't know where that value is coming from, so either a lot of investors are getting fleeced, or this is a Manhattan tier strategic project... privately funded, which makes even less strategic sense.
It's fascinating how most people still don't get it.
ASI is basically a god. This is the ultimate solution (or problem). It will push us to the singularity, and create an utopia or drive humanity to extinction. Imagine someone who is so smart that would win every single nobel prize available, and make multiple discoveries in a matter of a year. And now multiply this person's intelligence by 100 (most likely more, but 100 is already hard enough to grasp). There's no point in investing in anything else. An investment in ASI is an investment in everything (could be a bad one though, depending on the outcome).
The government is banking on being able to control it, which is also pretty funny. It's like a pet hamster thinking they can dictate what a human does.
I'm not sure I buy the national security argument but as you say the other side can trivially steal or absorb theoretical discoveries but not trivially get $500bn worth of data centers.