CapEx Full-Year 2024: $39.23B
Expected to increase to $60B-$65B in 2025, largely to support AI efforts.
NVDA bumped after hours on this news about 1.6% at the moment after a down day
It’s kind of like when 3d graphics came out. We all kind of knew ‘the graphics are good, but not there yet’.
Eventually it does get to Crysis level graphics, and we all take a breather. But then it starts up again with Unreal 5 engine. We will get to photorealistic even if takes a 1000 years.
Same goes for AGI.
Photorealism is basically a straight shot, the algorithms for light are well understood, we just haven't had the hardware to do it in realtime, but the hardware was relatively easy to iterate on.
AGI is not the same. No, LLMs are not going to turn into an "Artificial General Intelligence" just because we keep improving them - LLMs inherently don't have the capability of thought, interest, reason, motivation, or anything we ascribe to human intelligence. And throwing faster hardware at the problem isn't going to solve that.
Sure, people are going to keep trying but I doubt anyone alive today will be seeing a true artificial general intelligence. In 1000 years? That's quite the goalpost.
https://en.wikipedia.org/wiki/Artificial_general_intelligenc...
A truly intelligent machine would need the capability to reason, understand, and think, exhibiting cognition, sentience, and consciousness - not merely sift through training data for a plausible response to an input. That isn't intelligence, that's a fancy search engine.
LLMs and the entire premise behind them is incapable of producing AGI. They might seem like "AI", but they are nothing at all like AGI.
No, we certainly do not - we only have a very basic understanding of the biological component, nowhere near enough to replicate consciousness. You're fooling yourself if you think it's just a matter of throwing hardware at the problem.