원문 보기
the first part of what Sac said I think is really important for entrepreneurs to internalize which is where can we make money open AI has taken the Internet by storm in the past month and many across traditional media and social media have been speculating that For Better or For Worse no industry is safe from being affected crypto included some are predicting that open AI can bring significant benefits to the crypto industry such as increased efficiency improved security and compliance while some are predicting that open AI will lead to better art and narration in web 3 games open AI made headlines this week as Microsoft made a 10 investment into openai in a complicated deal which if all goes through looks as if Microsoft would have a 49 stake in the company openai started off as a not-for-profit project as it was deemed too powerful to be in the hands of one private company That Tune changed once the creators realized it was too powerful to be left to the open public pundits in analysts fear that openai will take out the workforce and not just entry-level jobs but white collar traditionally high-paying jobs investor shamat palihapatia has a different view on what's possible in his all-in podcast shamat compared openai to that of ingredients at a grocery store he believes that the real opportunity is going to be for those entrepreneurs and investors who know how to take the ingredients and create their own special meal and instead of looking at open AI from a fearful perspective look at it as a real opportunity to make serious money let's listen to shamath explain the reality is that well let me just take a prediction I think that Google will open source their models because the most important thing that Google can do is reinforce the value of search and the best way to do that is to Scorch the Earth with these models which is to make them widely available and as free as possible that will cause Microsoft to have to catch up and that will cause Facebook to have to really look in the mirror and decide whether they're going to cap the betting that they've made on AR VR and reallocate very aggressively to AI I mentioned this in the I did this Lex Friedman podcast but that should be what Facebook does and and the reason is if Facebook and Google and Microsoft have roughly the same capability in the same model there's an element of machine learning that I think is very important which is called reinforcement learning and specifically it's reinforcement learning from Human feedback right so these rlhf pipelines these are the things that will make your stuff unique so if you're a startup you can build a reinforcement learning pipeline how you build a product that captures a bunch of usage we talked about this before that data set is unique to you as a company you can feed that into these models get back better answers you can make money from it Facebook has an enormous amount of reinforcement learning inside of Facebook every click every comment every like every share Twitter has that data set Google inside of Gmail and search Microsoft inside of Minecraft and Hotmail so my point is David's right the huge companies I think will create the substrates and I think they'll be forced to Scorch the Earth and give it away for free and then on top of that is where you can make money and I would just encourage entrepreneurs to think where is my Edge in creating a data set that I can use for reinforcement learning that I think is interesting that's kind of saying I buy the ingredients from the supermarket but then I can still construct a dish that's unique and you know the salt is there the pepper is there but how I use that will determine whether you like the thing or not and I think that you know that is the way that I think we need to start thinking about it the biggest opportunity here is for Facebook I mean they need to get in this conversation ASAP I mean to think that like look Pi torch was like a pretty seminal piece of technology that a lot of folks in in Ai and machine learning were using for a long time tensorflow before that and what's so funny about like Google and Facebook is they're a little bit kind of like they're not really making that much progress I mean Facebook released this kind of like Rando version of alpha fold recently it's not that good I think these companies really need to get these products in the wild as soon as possible it cannot be the case is that you have to email people and get on some list I mean this is Google and Facebook guys come on by the way Jason this what you're doing right now is you're creating a human feedback reinforcement learning pipeline for chat GPT so just the fact that you asked that question and you know over time if chat GPT has access to your GPS information and then knows that you went to restaurant a versus B they can Intuit and it may actually prompt you to ask hey Jason we noticed you were in the area did you go to Bottega if you did how would you rate it one through five that reinforcement learning now allows the next person that asks what are the top five restaurants to say well you know over a thousand people that have asked this question here's actually the best answer versus a generic rank of the open web which is what the first data set is that's what's so interesting about this so this is why if you're a company that already owns the eyeballs you have to be running to get this stuff out there I was thinking about what you said last week and I thought back to the music industry and in the world of Napster and what happened was there was a lot of musicians I think Metallica being the most famous One famously suing Napster because it was like Hey listen like you're allowing people to take my content which they would otherwise pay for there's economic damage that I can measure that legal argument was meaningful enough that ultimately Napster was shut down now there were other versions of that that folks created including us at Winamp we created a headless version of that but if you translate that problem set here is there a claim that Yelp can make in this example that they're losing money that you know if you were going through Google or if you're going through their app there's the sponsored link revenue and the advertising Revenue that they would have got that they wouldn't get from here now that doesn't mean that chat GPT can't figure that out but it's those kinds of problems that are going to be a little thorny in these next few years that have to really get figured out just to build on what you said this week is the JPMorgan conference Friedberg mentioned it last week I had dinner on Wednesday with this really interesting company based in Zurich and what they have is basically a library of ligands right and so these ligands are used as a substrate to deliver all kinds of molecules inside the body and what's interesting is that they have a portfolio of like a thousand of these but really what they have is they have all the nuclear medicine about whether it works so you know they're they target glioblastoma glioblastoma and so all of a sudden they can they can say well this ligand can actually cross The Blood Vein barrier and get to the brain they have an entire data set of that and a whole bunch of nuclear imagery around that they have something for Soft Cell carcinoma so then they have that data set so to your point that's really valuable because that's real work that Google or Microsoft or open AI won't do right and if you have that and you bring it to the problem you can probably make money you know there's a business there to be built to watch the full video check out the link in the description what do you think of shamat's thoughts here what do you think about openai and how will it affect the crypto industry leave a comment if you found value in the content please consider donating to our small Channel by clicking the thanks button below that's it for today guys thanks for watching and we'll see you in the next one
Comments 0