正在加载视频...
视频加载失败
Mathlib is Lean library that has formalized a lot of humanity's math concepts and proofs. Grant speculated about a really interesting possibility: You could just have different AIs fork MathLib and build their own parallel math civilization. Lean guarantees that whatever new conjectures they come up with would be... show more
0 条评论
暂无评论
原始帖子的评论将显示在这里
![Naval Ravikant’s checklist for starting a company “The most important thing is there are no formulas. At the end of the day, you have to do what you love, and you have to do it even though people tell you it’ll never work. But that being said, if there was a formula [for starting a company], I would put it something like this.” Naval started seven companies before AngelList and this is the checklist he recommends running through before starting a startup: 1. Pick a great cofounder. This is most important: “You can do a company on your own, but it’s like you can raise a child on your own, but you probably shouldn’t. You need someone who’s going to be there with you.” This has it’s own checklist. Your cofounder should be: a. Very high intelligence (”hopefully they make you feel dumb, or they’re not smart enough”) b. Very high energy (”They should be extremely hardworking. A founder is someone who never has to be motivated. You should not have to be telling them to do their job.”) c. Very high integrity. (”a smart, hardworking crook who’s going to cheat you is the worst kind of person to be paired up with.”) 2. Pick a very large market. “Notice I don’t talk about the idea. I think ideas are almost irrelevant… The more important thing is that you pick a large space that you’re knowledgeable and passionate about. And then you will figure out what the right thing to do within that space is.” You want to be able to say to investors: “This is a space where there’s a huge market. I’m really knowledgeable and passionate about it. Here’s the great person that I have doing it with me. And here’s the minimum viable product that we have built. That will show that we can test in the marketplace… You iterate until you get to product/market fit… And then you go and you raise money from people you trust. And you use that money to scale.”](https://image.24vids.com/tw-1930592783535874225/amplify_video_thumb/1930575280118091776/img/tckUZIL13jo_NRF_.jpg)
!["One of the very confusing things about the models right now: how to reconcile the fact that they are doing so well on evals. And you look at the evals and you go, 'Those are pretty hard evals.' But the economic impact seems to be dramatically behind. There is [a possible] explanation. Back when people were doing pre-training, the question of what data to train on was answered, because that answer was everything. So you don't have to think if it's going to be this data or that data. When people do RL training, they say, 'Okay, we want to have this kind of RL training for this thing and that kind of RL training for that thing.' You say, 'Hey, I would love our model to do really well when we release it. I want the evals to look great. What would be RL training that could help on this task?' If you combine this with generalization of the models actually being inadequate, that has the potential to explain a lot of what we are seeing, this disconnect between eval performance and actual real-world performance"](https://image.24vids.com/tw-1993450075616690474/amplify_video_thumb/1993449024045932544/img/W-1TjOXcXeiUbHaH.jpg)
