@elonmusk Reminds me of the trans issue, first they denied surgery was being performed on minors then when the evidence came out, they said it was a good thing!
I'm not sure Elon really cares. His job is to put as much pressure as possible on his team and they keep making him look bad, which allows him to keep more pressure on to deliver something that's extremely difficult. Whether he contains the message the way you want him to do it, I think, has no bearing on the fundamentals of the business. Either they're going to deliver on time or they are not.
The trillions in AI capex are inference-driven. Training the next big model is in the 10s of billions?. We could fund that without closed-model profits, then let others run the services. Might take longer but we would get there irrespective.
The big labs have monopoly ambitions (god complex) and that’s why you see them rushing to capture everything. Not sure it’s a good idea for the intelligence maker to control so much, I think it would be healthier for them to have calmer ambitions.
@Suhail They are distilling, that’s mainly what they have been doing for the last 30 years. Ripping off western tech and making it more efficient and cheaper.
@iamgingertrash Mate we are entering the age of abundance and your crying about style…. It is inevitable and the US will get credit for it either way. Cultural victory was won in 1776
Still, in the end, open-weight models deter further AI capex.
Yes and no, the evidence so far shows no but conceptually when the open models are good enough it certainly could slow capex spend down.
However, it is worth to seperate out what part of capex spend actually decels. Training capex relative to inference is widening by the day. I do not belive that even if you get par open models that labs will slow training capex because it’s not actually that much in the grand scheme of things.