저자: ChamathPalihapitiya · 외국인
요약 보기
로봇/피지컬 AI가 본격화되고 있으며 Figure의 Brett Adcock가 가정용 로봇 베타 테스트 시점을 앞당겼다고 언급. 다만 범용 AI 모델과 액추에이터의 정밀도가 아직 한계이며 실용화까지 수년 걸린다고 봄. 엔비디아는 데이터센터 매출이 전년 대비 78% 증가하고 피지컬 AI 협업을 확대 중이며, 엣지 컴퓨팅 수요도 주목. 스타트업 투자는 AI로 인해 유사 아이디어 경쟁이 치열해져 승자 선별이 어려워진다고 언급.
다듬은 글
anyways did you guys see Brett atc's tweet no what did he say friend of the Pod Brett atock who's the CEO and founder of figure he just announced today that he's moved his timelines up by two years he's going to Beta Testing robots in the home by by the middle to end of this year that's crazy crazy crazy I mean are you an investor in his company I don't talk about my investments but in this case here but in this case no I'm not an investor so we're not talking a book here I do think Optimus and this figure and the other this about a dozen of these doing it credibly if they can make these for what chth 20 grand when do you think it becomes something a middle class household you know dual income household would buy one of these five years from now 10 years from now how how could they be well I think the issue is bounded by two things one is that I'm not sure that the generalized AI is good enough yet and what he did was he had a deal with open AI which he pretty publicly canell a few weeks ago and he announced his own model and I don't know the details of it to know whether he rolled it himself or this is just like taking some open source based model and iterating from it but I think the model is not perfect yet to be general purpose that's one and then the second is a practical issue with the robots which is that the actuators themselves are good but they're not great and you can see it in in the demo where it's an incredible demo because it shows the value and the power of the model where there's sort of this Master Slave orientation that has to happen where one model is actually doing most of the computation and the second model and the second robot is then feeding off of it and the demo that they do Nick you can probably find the video is of them sorting a bag of groceries for the first time totally unsupervised which it's it's an incredibly cool demo it's a cool demo the thing that that you notice though is that the actuators are good they're not great and so the physical dexterity is still relatively limited and I think that that doesn't allow these robots to be super functional in the next couple of years but when they get that figured out then I think it could be really useful because if you have a robot like this that could sort the groceries make food do the laundry mow the lawn so to speak it just requires level of dexterity that's not yet totally possible but see in this example what you're seeing are the two robots basically figuring out how to communicate semantically between the robots and that's incredibly powerful and it's yet another sort of breakthrough that we need so I don't know we're probably like a couple years away but see look at the dexterity there he's taking the pepp farms and feels like he's crushing the pepp Farms or she or whatever you call this robot please don't misgender the robot but it's it's really incredible they're figuring the the coolest part of this demo by the way which I loved was they take an apple and then the second robot figures out that it should go in the fruit bow pushes the fruit bow to the first robot and then the first robot that's cool there it is but that level of semantic awareness and understanding between two models working dependently is very cool chth piapa recently came out to talk about the unpredictable future of Robotics while dropping some Bombshells about how Nvidia is progressing nvidia's CEO Jensen hang recently teased this groundbreaking vision during their buzzworthy earnings call though the company has set the tech sphere a light with its thriving Data Center business generating 78% more Revenue year-over-year it's nvidia's latest Endeavors with physical AI that have everyone's attention welcome back everyone to invest docy they're collaborating with each other makes total sense I can tell you here on the ranch I would love to have an allpurpose robot going out there and using the weed whacker and trimming the the the the bushes and the hedges and getting me wood and collecting chicken eggs like there's a million things they could do on a ranch be immediately applicable for ranch work and if they're 24 hours a day it doesn't matter if they go slow but I I think this is the category people are sleeping on and I I don't know who on the on our prediction show said this will be the year of robots but it's been this is the year of robots for 30 years in the industry and it does feel like this is it freeberg yeah you're sticking with your prediction I assume yeah I am yeah I think it's also it's not just this kind of dexterous automation but I do think drones autonomous vehicles I put them all in the same category where there there's some it's some combination of mechanical response to a machine Vision system that uh I think is become like accelerated this year you need a lot of rare Earths to make robots yeah where we ever get those from hm does anybody owe us a little money is anybody behind on their payments maybe The Vig could be a little you know taste I don't know if you guys have seen this but there was a company in the 90s that was all the rage called segue they were going to absolutely change cities and everything and they never did but it was basically like a scooter you could walk on and it had a a balancing uh kind of system to it but they made robots and here they're making these lawnmowers now and these lawnmowers are really like cheap they're a th000 bucks and they work here in Austin I have seen two or three of these on people's Lawns this could be like you know what was the one that you did in your house Roomba so your point freeberg there will be Purpose Driven ones to deliver you a burrito do your lawn Etc and a room is like three or 400 bucks I think and this thing's a th000 bucks man this is going to get crazy I do think it's it's a lot harder to create one of these general purpose systems in automation then create like vertically or kind of utility specific automation system so a device that just does one thing delivers something to you in the air or drives your food to you or loads and unloads your dishes you know I'm not sure if that's the whole the whole idea of the humanoid is it's ambitious yeah it is a general purpose device and it makes a a technically very hard road map I got to imagine some of the bulldozers out there are also now becoming remote controlled so you can get like a a bulldozer that you have to put in a dangerous situation and they're remote control and they're going to have ai so I got pitched on a startup one time that was going to go up into Tahoe Hills and allow humans with remote controls to drive J off little bulldozers and make fire pths so imagine some fire brakes out you send in or helicopter in the bulldozzer no human in it just got a 5G connection or a starlane connection and Zip Zip Zip you're doing fire roads in the middle of the smoke dense area it's going to be really interesting when these things uh get dialed in and they're getting dial in every day I'm gonna ask Brett to be a b Alpha tester of one of these robots in my house and then we'll go and do we'll do a segment that would be great Nvidia is not just staying afloat it's riding the crest of the AI wave as CEO Jensen hang addressed during their fourth quarter earnings call the current sensation surrounding AI agents is significant yet the real game changer lies ahead in the evolution towards physical AI this is not sci-fi this is the next Frontier that Nvidia along with collaborators like uber 1X and neuro robotics is eagerly tailoring through advanced World models in the cosmos platform the aim to craft realistic simulations vital for testing autonomous Technologies in varied scenariof you're taking in a lot of data and then you have to run a lot of data in a model it's a lot faster to to run that model locally like when Tesla runs self-driving it's not sending the video images from your car to a server a th000 miles away and then letting the server decide how to drive your car the car is running its model on what to do with respect to the video imagery in the car it's local because the ability for all that data to get processed in the car means that you don't have to wait for the internet to transmit data back and forth you don't have lag time you don't have the 60 millisecond or 100 millisecond response time you don't have it losing your phone connection and then not knowing what to do exactly or or the the connection drops or waiting for a server to come online or server breaks in the data center everything is local so if you strap this like you know Nvidia computer which is basically plug andplay you don't have to have like Hardware expertise you could strap it onto the back of a humanoid robot or run research applications locally so I think that there's going to be some really interesting use cases whether it becomes a replacement for the Apple you know Macintosh Pro Studio device whatever maybe we'll see Mac Mini 4 yeah the Mac Mini 4 but a lot of people have pointed out that actually the the compute on this thing for $3,000 knocks a lot of Maxs out of the um out of the field I just can't runting system in the traditional sense Sam when we look at startups I remember when you and I started investing two of the driving forces was free storage free bandwidth and cloud computing drove a lot of ability to get a product to Market very quickly effectively Etc what impact will AI have on all these startups that are being originating now in 2024 2025 look into your crystal ball and how do you think they'll uh grow the footprint of them how is this going to Exel at the startup scene I actually think we're going to see a Cambrian explosion of creativity and and development of different things and some of them are going to be stupid ideas and some of them are going to be great but I think it's going to make our job especially at the seed stage of investing harder and harder there's going to be so many there's just going to be a lot of people that have similar ideas at the same time that can execute quickly and and do things that break next speeds that they've never been able to do before and um you know picking the winner is going to be hard to figure out it's going to be harder and harder yeah and it might be that um you know I've been thinking about this like do you in invest in competitors which is something I never used to do you know do you take a bet an index an entire category that you're interested in you know what is the Approach at seed And precede because uh I think of an idea and I'm like wow that's really neat and then I go and look out there and there's 30 people working on it and that didn't used to be the case and I think part of it is uh We've really unlocked a tool that allows people to do things that would have been cost prohibitive or gives them the ability to think gosh I could be an entrepreneur and I can try this and I could do this so I'm seeing people experiment and do all sorts of things as far as the startups some of AI stuff is just a feature you know it's just um table Stakes at this point it's like you know a chat or whatever and that doesn't really matter but then you're seeing people reimagine games and reimagine you know even things down to your kitchen appliance Etc uh so I just I do think it's going to be very very difficult and and I tend to set out a lot of hype Cycles so I invested in power and compute lithography kind of all of the things that are going to be underneath all of this and so I'm not sure how much of it I'm going to participate in until it starts to get to a steady state and you kind of can understand what's next because the the rate of acceleration is just so great that it's just kind of unclear to me sometimes you know especially when it comes to these consumer applications consumer facing things it's just uh really hard going Beyond traditional AI that powers chatbots and virtual assistance physical AI will grasp the nuanced dynamics of our tangible environment imagine self-driving cars seamlessly navigating diverse weather conditions or robots performing delicate tasks safely thanks to nvidia's pioneering developments these technological Feats are bolstered by promising numbers with Nvidia witnessing a commendable 78% Revenue increase to $ 39.3 billion and an 80% jump in profit largely attributable to its thriving data center Venturen we were picking famously Uber you had to pick between scar lift and Uber were three people doing it and it was pretty clear who was the most qualified you know amongst those three you now to your point if you want to be involved in Tax Plus AI or legal plus AI you might be looking at 50 companies a 100 and it was tradition in Silicon Valley to not bet on competitors there were some notable exceptions when you run an accelerator like I do Tech Stars or Y combinator you aren't Bound by that because 50% of the companies pivot almost by Design so I think you just have to I think at preced because people pivot you just have to tell people like listen we have a lot of pivoting going on people are going to run into each other I can't just bet on one thing right you in a space but I think that's a reasonable compromise if if all the founders are going to keep pivoting to each other's businesses how can the investors even keep track of that it's like being aircra traffic control of 10 airports at once it's just not feasible uh let's you go ahead you wouldn't think it but there's still a lot of spreadsheet companies out there you know you think you'd run out of them but they're still out there you know you look at and I think this is where AI is really going to make a difference like RFP proposals for governments you know something that takes like 30 days and it's manual and you have to submit these horrible documents you know you can ingest your entire Corpus of you know all of your previous bids and submit them at a break neck speed now and win more contracts that becomes like a National Defense company at that point and and so I think we're going to see a lot of really interesting things where a lot of Cru is going to disappear and and that'll be a really interesting wave that I'm I'm looking forward to yeah in fact jamaath has made a big bet there with his time uh with his uh software startup that he's created in the rapidly advancing realm of physical AI Nvidia stands poised as a leader transforming Visions into tangible realities as it envelops Robotics and Automotive Innovations with groundbreaking simulations the prospect of a world powered by intelligent autonomous Technologies is no longer a distant dream
Comments 0