Note that the LLVM approach to IR is probably a bit more sane than the GCC one. GCC has ~3 completely different IRs at different stages in the pipeline, while LLVM mostly has only canonical IR form for passing data around through the optimization passes (and individual passes will sometimes make their own temporary IR locally to make a specific analysis easier).
the further back you go, the less evidence we have. we don't know it was this week 1000 years ago, it's just that the error bars got big enough that we can't yet rule it out.
those gross profit margins aren't that useful since training at fixed capacity is continually getting cheaper, so there's a treadmill effect where staying in business requires training new models constantly to not fall behind. If the big companies stop training models, they only have a year before someone else catches up with way less debt and puts them out of business.
Only if training new models leads to better models. If the newly trained models are just a bit cheaper but not better most users wont switch. Then the entrenched labs can stop training so much and focus on profitable inference
Well thats why the labs are building these app level products like claude code/codex to lock their users in. Most of the money here is in business subscriptions I think, how much savings would be required for businesses to switch to products that arent better, just cheaper?
> A Starlink satellite uses about 5K Watts of solar power. It needs to dissipate around that amount (+ the sun power on it) just to operate.
This isn't quite true. It's very possible that the majority of that power is going into the antennas/lasers which technically means that the energy is being dissipated, but it never became heat in the first place. Also, 5KW solar power likely only means ~3kw of actual electrical consumption (you will over-provision a bit both for when you're behind the earth and also just for safety margin).
This isn't a problem for evolutionary theory. It's literally a necessary prediction of it. Most recent common ancestor of humans and chimps is 5-10 million years ago. Since we have observed tool usage in modern chimps and lots of very complicated tool use in humans, the necessary prediction is that some amount of tool use goes back at least ~5-10 million years, with increased complexity roughly tracking with the continuous increase in braincase size.
if it were only 2 primates that's a plausible explanation, but when it's pretty much every simean using tools, and all the old world apes making tools, it's pretty hard to argue for convergent evolution rather than a trait that exists ancestrally.
reply