@fabrice_mayrand@Vlisco2020 You can't run inference on "all the latest models" if you don't have those models.
You could offer people money to steal their trajectories, and then sell those for distillation, and position that as a discount. Or you could lie and give them a worse model than you promised.
@fabrice_mayrand@Vlisco2020 This is 100% a scam. The only way you can earn money selling tokens for models you don't own is by selling data, serving worse models, or doing something sketchy
@VladChituc Time is the main limiting factor at big labs, not money. If you could reliably get a year's worth of progress done in a day by spending 365x as much, everyone would.
@VladChituc If you want to know what will happen next [TIME UNIT], your x axis should be TIME.
Your chart shows what will happen over the next trillion dollars invested. Which is meaningless, because if AI hits a wall, spending on it will instantly go to zero.
@awe0_io Its authors are great engineers! They built some of the best internal UI tooling I have ever seen, and Astryx is the open-source version of that!
@MatthewBerman If you could earn money by mindlessly buying tokens, then anyone with an internet connection, including whoever is paying you, could do the same thing.
You provide no real value -> the market gives you no margin
@alexandr_wang The last 4 months in MSL have been the most fun, demanding, and meaningful work I've done in my career. I've gotten to work with some of the smartest and kindest people I've ever met, who deeply care about what they build. Truly an honor to be here.
See you all in the future <3