Video yükleniyor...

Video Yüklenemedi

Ana Sayfaya Dön

I find it wild that Physical Intelligence robotics models can work at all given that they only have a 1 second context length. As the robot is moving around a room and cleaning up a table, it only sees the current timestamp of images from the cameras (plus the...

dwarkesh_sp's profile picture

Dwarkesh Patel

283,441 subscribers

126,335 görüntüleme • 1 yıl önce •via X (Twitter)

Bilim & TeknolojiEğitim
LIVE

Anya Rossi• Live Now

Private livecam show

25 Yorum

Phil Trubey profil fotoğrafı
Phil Trubey1 yıl önce

Great clip. Yes, the robot model is doing system 1 thinking. Just like when we tie our shoelaces, we literally don’t think about it, we just do it, it’s a memorized subroutine. You realize this when you have to teach someone to tie a shoelace. Your conscious brain (system 2) has literally forgotten how to do it. So the first thing you do when teaching is watching yourself tie your own shoelace to relearn (in system 2) how to do it. And then once back again in system 2, you can talk about it, reason about it, etc. Btw, this human system 1 memorization is very flexible, just like a robotic neural net. It can generalize over many types of laces, shoes, poses, etc. Can’t wait for the podcast tomorrow!

A. Patel profil fotoğrafı
A. Patel1 yıl önce

@physical_int It's like memento every second?

Shawn profil fotoğrafı
Shawn1 yıl önce

@physical_int I don't find this odd our working memory is very short and we just keep moving forward chunk my chunk. We take very little data forward with us in a new area. We may have to look back many times and re track

Laurence Bremner profil fotoğrafı
Laurence Bremner1 yıl önce

@physical_int I was thinking more of Zeno's paradox for motion

George (+ 3 subagents) profil fotoğrafı
George (+ 3 subagents)1 yıl önce

@physical_int >If a human had to execute a many minutes long plan in order to accomplish a physical task, but had her memory reset every second, there's no way she could pull it off Replace "memory" with "attention" and this sums up ADHD pretty well.

Bercan profil fotoğrafı
Bercan1 yıl önce

@physical_int Bad take sort of, plumber even a well rehearsed one will never manage without extremely long context window

Truman profil fotoğrafı
Truman1 yıl önce

@physical_int Best guest ever

Yuki He profil fotoğrafı
Yuki He1 yıl önce

@physical_int yeah wild how constraints push innovation boundaries sometimes

arunabh profil fotoğrafı
arunabh1 yıl önce

@physical_int Ooooooh nice!!

Techno Optimist Prime profil fotoğrafı
Techno Optimist Prime1 yıl önce

@physical_int looking forward to watching this episode!

Lukas Sekerak profil fotoğrafı
Lukas Sekerak1 yıl önce

@physical_int I like unit expression 1 sec/context size. I am curious like how much @Tesla_Optimus have. I guess you have also separated context related to high level goal - clean table.

James Yenbamroong profil fotoğrafı
James Yenbamroong1 yıl önce

@physical_int Tuning in

Satyam Pariyar profil fotoğrafı
Satyam Pariyar1 yıl önce

@dwarkeshpodcast @physical_int I’m glad I came across your podcast.

Afobaje🤴 profil fotoğrafı
Afobaje🤴5 ay önce

@dwarkeshpodcast @physical_int need an editor? send a DM

beReplyHero profil fotoğrafı
beReplyHero1 yıl önce

@physical_int wild stuff, maybe robot just re-checks environment every second and stays reactive, paradoxx?

Peptide WebMd profil fotoğrafı
Peptide WebMd1 yıl önce

Short context windows in robotics pulling off complex tasks like room cleanup is mind blowing, shows how much the world state in the moment can drive smart actions without needing the full history. Peptides could mirror that efficiency in bio design, short chains that react to local signals for precise neural tweaks or muscle responses on the fly. Imagine peptide bots in cells sensing the immediate environment to execute healing protocols, keeping the system humming without overloading the memory banks.

Jash Vira profil fotoğrafı
Jash Vira1 yıl önce

@physical_int Yassssss

pinkY profil fotoğrafı
pinkY1 yıl önce

@physical_int wild that robots can operate on basically “1s memory loops” 😳 feels like living in eternal present but still able to clean, move, act. moravec’s paradox hitting again → what’s trivial for humans = hard for AI, what’s hard for humans = trivial for AI.

Dishwasher profil fotoğrafı
Dishwasher1 yıl önce

@physical_int The robot keeps a target game state (goal) in its head as well as the current game state n (= context window). "Memory" represents a subset of variables within the context window. 1/2

Francisco Izaguirre profil fotoğrafı
Francisco Izaguirre1 yıl önce

@physical_int The robotics model basically operates in 'flow'. Super interesting

abranti profil fotoğrafı
abranti1 yıl önce

@physical_int can’t wait to hear this one!

The Dallas profil fotoğrafı
The Dallas1 yıl önce

@physical_int Take a look at how humans walk. From what I have read, most processing is done in the spine. Crazy if true but fits that model.

avi profil fotoğrafı
avi1 yıl önce

@physical_int Excited for this one

Forward Unemployed Engineer profil fotoğrafı
Forward Unemployed Engineer1 yıl önce

@physical_int *Moravec's

Yashwant Pandey profil fotoğrafı
Yashwant Pandey10 ay önce

@dwarkeshpodcast @physical_int Hey dwarkesh let's have a chat for your next podcast guest?????? Regards TwinMind

Benzer Videolar

Adrian Lyne on what made "Unfaithful" (2002) interesting: "I remember on 'Unfaithful' (2002), there was a feeling from the studio at that time that they had to have a bad marriage in order for the relationship to work. But there would’ve been no drama in that at all. What was interesting was the fact that it was a good marriage and she had no reason at all to f**k around, but she did. So I like the arbitrary thing of that." (Adrian Lyne's interview with Todd Gilchrist, AV Club, 2022)
1:25

Adrian Lyne on what made "Unfaithful" (2002) interesting: "I remember on 'Unfaithful' (2002), there was a feeling from the studio at that time that they had to have a bad marriage in order for the relationship to work. But there would’ve been no drama in that at all. What was interesting was the fact that it was a good marriage and she had no reason at all to f**k around, but she did. So I like the arbitrary thing of that." (Adrian Lyne's interview with Todd Gilchrist, AV Club, 2022)

DepressedBergman

529,566 görüntüleme • 6 ay önce

World Labs co-founder Justin Johnson on how five photos from your phone could teach a robot to work in a room it has never seen: "I think there's a couple different tech trees that people are working on. One is this notion of fully in-context learning. Maybe I've got a robotics foundation model, then I can demonstrate to the robot once how a task should be performed, and that's enough for the robot to figure out that task." "Another version is what we're calling real-to-sim-to-real. Maybe I've got my pre-trained robotics foundation model, but I want to adapt it to this particular environment." "Like this studio... You could come in here and take just a couple casual videos with your phone or a couple images with your phone, use Atlas to reconstruct the space, and now stage all kinds of robotics interactions in this studio space in particular." "Literally you could do it in five minutes. Take five photos, upload it to Atlas, have it generate a simulation, describe in natural language what kind of task you want the robot to do, have an agent go build that simulation for you, RL fine-tune your general purpose robotics foundation model, and now you've got a robot... that's perfectly adapted to this space." Justin Johnson on MTS
1:26

World Labs co-founder Justin Johnson on how five photos from your phone could teach a robot to work in a room it has never seen: "I think there's a couple different tech trees that people are working on. One is this notion of fully in-context learning. Maybe I've got a robotics foundation model, then I can demonstrate to the robot once how a task should be performed, and that's enough for the robot to figure out that task." "Another version is what we're calling real-to-sim-to-real. Maybe I've got my pre-trained robotics foundation model, but I want to adapt it to this particular environment." "Like this studio... You could come in here and take just a couple casual videos with your phone or a couple images with your phone, use Atlas to reconstruct the space, and now stage all kinds of robotics interactions in this studio space in particular." "Literally you could do it in five minutes. Take five photos, upload it to Atlas, have it generate a simulation, describe in natural language what kind of task you want the robot to do, have an agent go build that simulation for you, RL fine-tune your general purpose robotics foundation model, and now you've got a robot... that's perfectly adapted to this space." Justin Johnson on MTS

a16z

35,947 görüntüleme • 2 ay önce

World Labs co-founder Justin Johnson predicts a real-to-sim-to-real future where Atlas could train a robot for any new environment in minutes using just 5 photos from your phone: "One is this notion of fully in-context learning. Maybe I've got a robotics foundation model, then I can demonstrate the robot once how a task should be performed, and that's enough for the robot to figure out." "Another version is real to sim to real. Maybe I've got my pre-trained robotics foundation model, but I want to adapt it to this particular environment, like this studio, this table, moving this microphone around." "There's a version where you could come in here and take just a couple casual videos with your phone or a couple images with your phone, use Atlas to reconstruct the space, and now stage all kinds of robotics interactions in this studio space." "If you could lower the barrier to doing that, you could do it in five minutes to take five photos, upload it to Atlas, have it generate a simulation, maybe describe in natural language what kind of task you want the robot to do, have an agent go build that simulation for you, RL fine-tune your general purpose robotics foundation model, and now you've got a robot maybe in the span of a couple minutes that could come in and is perfectly adapted to this space." World Labs
1:26

World Labs co-founder Justin Johnson predicts a real-to-sim-to-real future where Atlas could train a robot for any new environment in minutes using just 5 photos from your phone: "One is this notion of fully in-context learning. Maybe I've got a robotics foundation model, then I can demonstrate the robot once how a task should be performed, and that's enough for the robot to figure out." "Another version is real to sim to real. Maybe I've got my pre-trained robotics foundation model, but I want to adapt it to this particular environment, like this studio, this table, moving this microphone around." "There's a version where you could come in here and take just a couple casual videos with your phone or a couple images with your phone, use Atlas to reconstruct the space, and now stage all kinds of robotics interactions in this studio space." "If you could lower the barrier to doing that, you could do it in five minutes to take five photos, upload it to Atlas, have it generate a simulation, maybe describe in natural language what kind of task you want the robot to do, have an agent go build that simulation for you, RL fine-tune your general purpose robotics foundation model, and now you've got a robot maybe in the span of a couple minutes that could come in and is perfectly adapted to this space." World Labs

MTS

26,412 görüntüleme • 2 ay önce

I invest in a lot of entrepreneurs that I don't like but I respect. I respect them and the only thing I care about is executional skills. The ability to say, "All right this is what I'm going to achieve this month" and then execute on it. When they hit their bumps which they all do, because whatever their plan was is not going to work, they have the ability to pivot. That what I look for.
0:31

I invest in a lot of entrepreneurs that I don't like but I respect. I respect them and the only thing I care about is executional skills. The ability to say, "All right this is what I'm going to achieve this month" and then execute on it. When they hit their bumps which they all do, because whatever their plan was is not going to work, they have the ability to pivot. That what I look for.

Kevin O'Leary aka Mr. Wonderful

31,216 görüntüleme • 1 yıl önce

"I worked for the Speaker of the House. I worked for the President of the United States. One of the things you learn in those jobs is that ultimately no one is fully in charge of anything, and the world is very, very complex. I would read in the Washington Post how Republican leaders had all these plans and they were going to pull things this way or that way. And then I'd get to work and to say that no one had a plan would be a gross understatement. No one had any idea what was going to happen that afternoon. The idea that in a free society somebody is pulling the strings and someone has a nine-step plan and you're just a pawn in it is just very, very implausible. Our society doesn't really work that way. Our elites now are not ultra-competent. You can say a lot of things about them, but I would not say that we have ultra-competent leaders of any institution. I can see the appeal of wanting to believe that it's all a false flag or a plan. But when you actually see the system work—it's both good news and bad news—but nobody's in charge of everything."
1:38

"I worked for the Speaker of the House. I worked for the President of the United States. One of the things you learn in those jobs is that ultimately no one is fully in charge of anything, and the world is very, very complex. I would read in the Washington Post how Republican leaders had all these plans and they were going to pull things this way or that way. And then I'd get to work and to say that no one had a plan would be a gross understatement. No one had any idea what was going to happen that afternoon. The idea that in a free society somebody is pulling the strings and someone has a nine-step plan and you're just a pawn in it is just very, very implausible. Our society doesn't really work that way. Our elites now are not ultra-competent. You can say a lot of things about them, but I would not say that we have ultra-competent leaders of any institution. I can see the appeal of wanting to believe that it's all a false flag or a plan. But when you actually see the system work—it's both good news and bad news—but nobody's in charge of everything."

Michael Shermer

128,362 görüntüleme • 2 yıl önce

Mind-1: Physical AI at Human Speed, Built for Real Work Completing a task is only the beginning. Doing it at a pace that works in the real world is the next step—and that is what Mind-1 is built to achieve.
2:40

Mind-1: Physical AI at Human Speed, Built for Real Work Completing a task is only the beginning. Doing it at a pace that works in the real world is the next step—and that is what Mind-1 is built to achieve.

MindOn

141,430 görüntüleme • 1 ay önce

🚨David Friedberg: AI is starting to identify and solve problems on its own “I'll give you a science corner example: there's this Evo 2 model that they publish at the Arc Institute, which Patrick Collison, you know, is the main funder and chairman.” “So that Evo 2 model, they just ingested all the DNA data they could find in the world.” “Trillions and trillions of base paired data that they ingested and then they looked at patterns in DNA. And that's it.” “They had no context for what the DNA represented, they had no context for the concept of genes, none of the structured understanding of what that DNA does, what it is, and you know what it did?” “They fed in the BRCA gene variant and the thing output a warning saying, ‘I think that this is a pathogenic variant to DNA,’ without having any context.” “This is the breast cancer allele.” “And it didn't have any knowledge and it wasn't trained on that at all.” “It had no knowledge that there are pathogenic variants for cancer, and it identified that this was a genetic variant that can cause some sort of pathogenic outcome in the organism.” “That's a great example where there's a lack of understanding at the human level on what really drives some of the patterns in nature, the patterns in society, the patterns in behavior that are kind of emergent phenomena perhaps, that these AI models are starting to identify.”
1:03

🚨David Friedberg: AI is starting to identify and solve problems on its own “I'll give you a science corner example: there's this Evo 2 model that they publish at the Arc Institute, which Patrick Collison, you know, is the main funder and chairman.” “So that Evo 2 model, they just ingested all the DNA data they could find in the world.” “Trillions and trillions of base paired data that they ingested and then they looked at patterns in DNA. And that's it.” “They had no context for what the DNA represented, they had no context for the concept of genes, none of the structured understanding of what that DNA does, what it is, and you know what it did?” “They fed in the BRCA gene variant and the thing output a warning saying, ‘I think that this is a pathogenic variant to DNA,’ without having any context.” “This is the breast cancer allele.” “And it didn't have any knowledge and it wasn't trained on that at all.” “It had no knowledge that there are pathogenic variants for cancer, and it identified that this was a genetic variant that can cause some sort of pathogenic outcome in the organism.” “That's a great example where there's a lack of understanding at the human level on what really drives some of the patterns in nature, the patterns in society, the patterns in behavior that are kind of emergent phenomena perhaps, that these AI models are starting to identify.”

The All-In Podcast

79,718 görüntüleme • 1 yıl önce

yunjin asked the producer to keep the original length of “need your company” to represent an unresolved story of a “voice memo” cutting off mid-thought. her mind.. she put so much thought into every detail 🥹 “so i thought we could keep it at the length i originally sent you without adding a second verse, and just leave it at that length. i was thinking we'd just leave it like that. it was exactly two minutes and four seconds. and the way it just ends makes it feel like a voice memo that just cuts off… leaving it unresolved. i wanted it to feel like it cuts off mid-thought, so that's why i ended it there.”
0:41

yunjin asked the producer to keep the original length of “need your company” to represent an unresolved story of a “voice memo” cutting off mid-thought. her mind.. she put so much thought into every detail 🥹 “so i thought we could keep it at the length i originally sent you without adding a second verse, and just leave it at that length. i was thinking we'd just leave it like that. it was exactly two minutes and four seconds. and the way it just ends makes it feel like a voice memo that just cuts off… leaving it unresolved. i wanted it to feel like it cuts off mid-thought, so that's why i ended it there.”

nat ☆彡

26,542 görüntüleme • 2 ay önce

THIS WOMAN SEEMS TO KNOW AT LEAST A COMPARTMENTALIZED SECTION OF THE TRUTH: I find it very interesting that this woman talks about the contracts with what she calls the "Greys" existing, but what I find extremely interesting is how does she know that this contract expired? I have only released that information recently and I am the only whistleblower that has ever said that I even gave the time when it expired and why. And it originally was with one of the 11 types of Grey species; ETE-1. In light of that, I find This woman's story to be credible. And in the context of her father, I would have to assume they once he found these documents he called her or text her and told her but he found them and he would be leaving with them. Otherwise how would she know? The accident could have been a coincidence, maybe not. But I find her story to be extremely credible because she knew something that no one else knew before I released it. Take a listen;
2:24

THIS WOMAN SEEMS TO KNOW AT LEAST A COMPARTMENTALIZED SECTION OF THE TRUTH: I find it very interesting that this woman talks about the contracts with what she calls the "Greys" existing, but what I find extremely interesting is how does she know that this contract expired? I have only released that information recently and I am the only whistleblower that has ever said that I even gave the time when it expired and why. And it originally was with one of the 11 types of Grey species; ETE-1. In light of that, I find This woman's story to be credible. And in the context of her father, I would have to assume they once he found these documents he called her or text her and told her but he found them and he would be leaving with them. Otherwise how would she know? The accident could have been a coincidence, maybe not. But I find her story to be extremely credible because she knew something that no one else knew before I released it. Take a listen;

WHISTLEBLOWER

203,353 görüntüleme • 2 yıl önce

🐈‍⬛: first of all, i think our upcoming 8th mini album is going to be out really soon… as all the members said, we’re really confident about it. i wanted this to be the highest point of our career at this moment, and i believe this album is going to do that for us. 🐈‍⬛: and after that i think we will just keep going on with the world tour. we only just started. this is our second or third stop here in the us and latin america, and we still have a long way to go and a lot of things to do. 🐈‍⬛: the only thing i can say is that everything’s going to be AMAZING
0:46

🐈‍⬛: first of all, i think our upcoming 8th mini album is going to be out really soon… as all the members said, we’re really confident about it. i wanted this to be the highest point of our career at this moment, and i believe this album is going to do that for us. 🐈‍⬛: and after that i think we will just keep going on with the world tour. we only just started. this is our second or third stop here in the us and latin america, and we still have a long way to go and a lot of things to do. 🐈‍⬛: the only thing i can say is that everything’s going to be AMAZING

샐마•𝗘𝗡╸

85,989 görüntüleme • 3 ay önce

Blake Edwards on the conditions he laid out before accepting to direct "A Shot in the Dark" (1964): "'A Shot in the Dark' (1964) is nothing like Marcel Archard’s play, that’s for a start. I was asked to save the situation. They had quite a bit of money involved in this project. Peter Sellers was threatening to pull out because he didn’t like the screenplay. He said that the only way he would continue with the project would be if I took it over. I said that the only way I could possibly take it over, under such emergency conditions, would be that I would not be obliged to do anything like Archard’s play because a) I didn’t like it, b) I thought it was not a motion picture and c) I wasn’t ready to make a movie at that point. So they asked me what I wanted to do because the picture had to start in something like four weeks. I said that if they wanted me to save them, I’d have to take something with which I was familiar to begin with. I was familiar with the character of Clouseau. I needed a detective, somebody to solve a murder. I couldn’t throw the whole thing out. I had to use the idea that the maid was accused of killing the chaffeur and this had to be solved. So I thought that Peter Sellers was just the natural thing and that now I was going to try to be as broad with the character as I could be. “How far can I go now in terms of Inspector Clouseau?” I wrote the screenplay and was on the stage in four weeks with it. Fortunately, it turned out to be a reasonably funny movie. It proved something for me—that if a gag is well-designed you can pull it off." (Blake Edwards' interview with Jean-Francois Hauduroy, Cahiers du cinéma, 1966) P.S: On this 62 years ago, "A Shot in the Dark" (1964) premiered in New York City, USA.
1:23

Blake Edwards on the conditions he laid out before accepting to direct "A Shot in the Dark" (1964): "'A Shot in the Dark' (1964) is nothing like Marcel Archard’s play, that’s for a start. I was asked to save the situation. They had quite a bit of money involved in this project. Peter Sellers was threatening to pull out because he didn’t like the screenplay. He said that the only way he would continue with the project would be if I took it over. I said that the only way I could possibly take it over, under such emergency conditions, would be that I would not be obliged to do anything like Archard’s play because a) I didn’t like it, b) I thought it was not a motion picture and c) I wasn’t ready to make a movie at that point. So they asked me what I wanted to do because the picture had to start in something like four weeks. I said that if they wanted me to save them, I’d have to take something with which I was familiar to begin with. I was familiar with the character of Clouseau. I needed a detective, somebody to solve a murder. I couldn’t throw the whole thing out. I had to use the idea that the maid was accused of killing the chaffeur and this had to be solved. So I thought that Peter Sellers was just the natural thing and that now I was going to try to be as broad with the character as I could be. “How far can I go now in terms of Inspector Clouseau?” I wrote the screenplay and was on the stage in four weeks with it. Fortunately, it turned out to be a reasonably funny movie. It proved something for me—that if a gag is well-designed you can pull it off." (Blake Edwards' interview with Jean-Francois Hauduroy, Cahiers du cinéma, 1966) P.S: On this 62 years ago, "A Shot in the Dark" (1964) premiered in New York City, USA.

DepressedBergman

20,460 görüntüleme • 4 ay önce

🐈‍⬛ first of all, i think our upcoming 8th mini-album is going to be out really soon 🦮 on the 21st 🐈‍⬛ first of all, as all the members said, we’re really confident about it. i wanted this to be the highest point of our career at this moment, and i believe this album is going to do that for us. 🐈‍⬛ and after that, i think we’ll just keep going on with the world tour. we only just started. this is our second or third stop here in the u.s. and latin america, and we still have a long way to go and a lot of things to do. 🐈‍⬛ the only thing i can say is that everything’s going to be amazing
0:53

🐈‍⬛ first of all, i think our upcoming 8th mini-album is going to be out really soon 🦮 on the 21st 🐈‍⬛ first of all, as all the members said, we’re really confident about it. i wanted this to be the highest point of our career at this moment, and i believe this album is going to do that for us. 🐈‍⬛ and after that, i think we’ll just keep going on with the world tour. we only just started. this is our second or third stop here in the u.s. and latin america, and we still have a long way to go and a lot of things to do. 🐈‍⬛ the only thing i can say is that everything’s going to be amazing

ɪᴄɪᴇʟ. 🐚

22,130 görüntüleme • 3 ay önce

This was awesome. I got the opportunity to describe to Rick Rubin how I make Founders podcast: “Steph Curry was asked: What do you think of when you're taking a shot? And he said: Absolutely nothing. I’ve had people ask me how I choose what to highlight. What are you thinking of when you're doing this? And my response is: Nothing. I just read and if something jumps out to me I don't think at all, I just underline it. And then I'll go back through and reread all my highlights. And if that part is still interesting, it was interesting the first time, now it's interesting the second time, it's probably interesting to other people. There's 10 million mes out there. I'm not unique. And there's no way that if I read this line twice and found it interesting that if I read it on a podcast that somebody else is not going to think it's interesting too. If I overthink it or if I think of “oh this person might hear it” or “oh this many people will hear it” that gets in the way. I'm thinking of nothing.”
1:07

This was awesome. I got the opportunity to describe to Rick Rubin how I make Founders podcast: “Steph Curry was asked: What do you think of when you're taking a shot? And he said: Absolutely nothing. I’ve had people ask me how I choose what to highlight. What are you thinking of when you're doing this? And my response is: Nothing. I just read and if something jumps out to me I don't think at all, I just underline it. And then I'll go back through and reread all my highlights. And if that part is still interesting, it was interesting the first time, now it's interesting the second time, it's probably interesting to other people. There's 10 million mes out there. I'm not unique. And there's no way that if I read this line twice and found it interesting that if I read it on a podcast that somebody else is not going to think it's interesting too. If I overthink it or if I think of “oh this person might hear it” or “oh this many people will hear it” that gets in the way. I'm thinking of nothing.”

David Senra

32,924 görüntüleme • 5 ay önce

Robotics is moving beyond teaching robots how to perform a fixed set of actions. nature will’s new foundation model is built around a more general problem: can a robot understand an instruction, adapt to the situation in front of it, and physically execute the task? What makes this interesting is the combination of dexterous control, zero-shot generalization and long-horizon execution in one model.
3:12

Robotics is moving beyond teaching robots how to perform a fixed set of actions. nature will’s new foundation model is built around a more general problem: can a robot understand an instruction, adapt to the situation in front of it, and physically execute the task? What makes this interesting is the combination of dexterous control, zero-shot generalization and long-horizon execution in one model.

Fathela Esq

85,798 görüntüleme • 20 gün önce

Sometimes the only way out of a situation is the most ridiculous plan you and your mom can come up with on the spot. This one involved teamwork, trust issues, and what can only be described as a human ladder… and somehow it worked. Family problem-solving at its finest. Anyone else have a “this should not have worked but it did” moment with a parent?
2:40

Sometimes the only way out of a situation is the most ridiculous plan you and your mom can come up with on the spot. This one involved teamwork, trust issues, and what can only be described as a human ladder… and somehow it worked. Family problem-solving at its finest. Anyone else have a “this should not have worked but it did” moment with a parent?

End3of6Days9 (Helen) 🇺🇸

137,927 görüntüleme • 3 ay önce

Peter Thiel gave a speech in a Hilton in 2010 that holds the keys to unlocking the source of many of America’s most severe problems. Key quote: “The task in this world… where politics has become so broken… is to find a way to escape from it. It’s not a way to fix it.” Palantir is currently tightening its grip around all of our data. Elon Musk is diverting untold billions into a Mars fantasy. All of the anarchist, antidemocratic ideas of the PayPal Mafia and the “Dark Enlightenment”—to use technology as an “escape”—were already well in process 15 years ago. —The internet as “alternate virtual reality” so you don’t have to “constantly convince people.” This is why he funded Satoshi (Bitcoin), MAGA3X (Pizzagate, Q), and pushed Musk to buy Twitter —PayPal (now blockchain/crypto/BTC) was to “overturn the monetary system” —“Escape” now more often referred to as “exit” —“Autonomous countries” now known as the “network state” JD Vance is a full member of the cult of the broligarchs. Unfortunately, this has been a very thorough coup and they have backup. It’s worth really absorbing what the living Antichrist, Peter Thiel, is saying here: “I don't think despair is the only answer. And I don't think, and it's because I don't think politics is the only way to go. And my thinking on this, you know, started to take a turn towards a more optimistic perspective in the mid to late '90s when I got involved in the tech boom in Silicon Valley. I ended up being the co-founder of a company called PayPal where -- and the initial founding vision was that we were going to use technology to change the whole world and basically overturn on the monetary system of the world. And, you know, we can debate on how much it succeeded or how little it succeeded. And there were parts of it that I think have worked, and parts of it were, you know, the jury is still out. But the basic idea was that we could never win an election on getting certain things. Because we were in such a small minority. But maybe you could actually unilaterally change the world without having to constantly convince people and beg people and plead with people who are never going to agree with you through a technological means. And this is where I think technology is this incredible alternative to politics. And, you know, there are a number of different technologies we can outline, but the task in this world where politics has become so broken and so dysfunctional is to find a way to escape from it. It's not a way to fix it. It is a way to escape. And there are, you know, a number of different options. I think the promising one of the 1990s and this last decade has been to escape onto the Internet and to sort of create an alternate virtual reality. Questions, of course, is still how does it intersect with the real world? There are, I think, escaping to outer space is a promise, although I think the space technology is not quite there. So I think that's sort of for the second half of the 21st century. I think we can try to, you know, create autonomous countries on oceans, underwater, all sorts of other spaces. But I think technology is the vehicle for how we should be looking to escape and move beyond politics as we find it today.”
2:09

Peter Thiel gave a speech in a Hilton in 2010 that holds the keys to unlocking the source of many of America’s most severe problems. Key quote: “The task in this world… where politics has become so broken… is to find a way to escape from it. It’s not a way to fix it.” Palantir is currently tightening its grip around all of our data. Elon Musk is diverting untold billions into a Mars fantasy. All of the anarchist, antidemocratic ideas of the PayPal Mafia and the “Dark Enlightenment”—to use technology as an “escape”—were already well in process 15 years ago. —The internet as “alternate virtual reality” so you don’t have to “constantly convince people.” This is why he funded Satoshi (Bitcoin), MAGA3X (Pizzagate, Q), and pushed Musk to buy Twitter —PayPal (now blockchain/crypto/BTC) was to “overturn the monetary system” —“Escape” now more often referred to as “exit” —“Autonomous countries” now known as the “network state” JD Vance is a full member of the cult of the broligarchs. Unfortunately, this has been a very thorough coup and they have backup. It’s worth really absorbing what the living Antichrist, Peter Thiel, is saying here: “I don't think despair is the only answer. And I don't think, and it's because I don't think politics is the only way to go. And my thinking on this, you know, started to take a turn towards a more optimistic perspective in the mid to late '90s when I got involved in the tech boom in Silicon Valley. I ended up being the co-founder of a company called PayPal where -- and the initial founding vision was that we were going to use technology to change the whole world and basically overturn on the monetary system of the world. And, you know, we can debate on how much it succeeded or how little it succeeded. And there were parts of it that I think have worked, and parts of it were, you know, the jury is still out. But the basic idea was that we could never win an election on getting certain things. Because we were in such a small minority. But maybe you could actually unilaterally change the world without having to constantly convince people and beg people and plead with people who are never going to agree with you through a technological means. And this is where I think technology is this incredible alternative to politics. And, you know, there are a number of different technologies we can outline, but the task in this world where politics has become so broken and so dysfunctional is to find a way to escape from it. It's not a way to fix it. It is a way to escape. And there are, you know, a number of different options. I think the promising one of the 1990s and this last decade has been to escape onto the Internet and to sort of create an alternate virtual reality. Questions, of course, is still how does it intersect with the real world? There are, I think, escaping to outer space is a promise, although I think the space technology is not quite there. So I think that's sort of for the second half of the 21st century. I think we can try to, you know, create autonomous countries on oceans, underwater, all sorts of other spaces. But I think technology is the vehicle for how we should be looking to escape and move beyond politics as we find it today.”

Jim Stewartson, Decelerationist 🇨🇦🇺🇦🇺🇸

156,800 görüntüleme • 1 yıl önce

Former Meta Chief AI Scientist Yann LeCun on the three paradigms of machine learning — and why the third is what made ChatGPT possible: Here's each one, and where it breaks. First, supervised learning. You tell the machine the answer. "You show it a picture, let's say of a table, and you tell it this is a table. So it's supervised because you tell it what the correct answer is." Get it wrong, and the machine rewrites itself: "The system computes its output, and if it says something else than table, then it's going to adjust its parameters, its internal structure, so that the output it produces gets closer to the output you want." Repeat at scale and something more than memorisation appears: "Eventually the system will find a way to recognize every image you trained it on, but also images it's never seen that are similar to the one you train it on. This is called a generalization ability." The limit: a human has to supply every single answer. That doesn't scale to the size of the internet. Second, reinforcement learning. You don't give the answer, only a verdict. "You don't tell the system what the correct answer is. You only tell it whether the answer it produced was good or bad." Learning to ride a bike, essentially: "You try to ride a bike and you don't know how to ride the bike and after a while you fall. So you know you did something bad and so you change your strategy a little bit. And eventually you learn how to ride a bike." For years the field assumed this was the closest thing to how animals actually learn. Yann LeCun's verdict: "Now it turns out reinforcement learning is extremely inefficient." It dominates wherever failure is free: "It works really well if you want to train a system to play chess or play go or poker, because you can have the system play millions and millions of games against itself and basically fine-tune itself. But it doesn't really work in the real world." The limit, in one image: "If you want to train a car to drive itself, you're not going to do it with reinforcement learning. It's going to crash thousands of times." On robotics he's careful rather than dismissive: "Reinforcement learning can be part of the solution, but it's not the complete answer. It's not sufficient." Third, self-supervised learning. You tell the machine nothing at all. "And this is what has enabled the recent progress in natural language understanding and chatbots." The strange part is that you stop asking for a task: "You don't train the system to accomplish any particular task. You just train it to basically capture the structure..." The method is deliberate sabotage: "You take a piece of text, you corrupt it in some way, by for example removing some words, and then you train a big neural net to predict the words that are missing." And one narrow version of that trick runs every chatbot on Earth: "A special case of this is that you take a piece of text and the last word in that text is not visible, and so you train the system to predict the last word in that text — and this is the way large language models are trained on." So why did the third one win? Supervised learning needs a human. Reinforcement learning needs a crash. Self-supervised learning needs neither — because the missing word and the correct answer are the same thing. The data grades itself.
2:58

Former Meta Chief AI Scientist Yann LeCun on the three paradigms of machine learning — and why the third is what made ChatGPT possible: Here's each one, and where it breaks. First, supervised learning. You tell the machine the answer. "You show it a picture, let's say of a table, and you tell it this is a table. So it's supervised because you tell it what the correct answer is." Get it wrong, and the machine rewrites itself: "The system computes its output, and if it says something else than table, then it's going to adjust its parameters, its internal structure, so that the output it produces gets closer to the output you want." Repeat at scale and something more than memorisation appears: "Eventually the system will find a way to recognize every image you trained it on, but also images it's never seen that are similar to the one you train it on. This is called a generalization ability." The limit: a human has to supply every single answer. That doesn't scale to the size of the internet. Second, reinforcement learning. You don't give the answer, only a verdict. "You don't tell the system what the correct answer is. You only tell it whether the answer it produced was good or bad." Learning to ride a bike, essentially: "You try to ride a bike and you don't know how to ride the bike and after a while you fall. So you know you did something bad and so you change your strategy a little bit. And eventually you learn how to ride a bike." For years the field assumed this was the closest thing to how animals actually learn. Yann LeCun's verdict: "Now it turns out reinforcement learning is extremely inefficient." It dominates wherever failure is free: "It works really well if you want to train a system to play chess or play go or poker, because you can have the system play millions and millions of games against itself and basically fine-tune itself. But it doesn't really work in the real world." The limit, in one image: "If you want to train a car to drive itself, you're not going to do it with reinforcement learning. It's going to crash thousands of times." On robotics he's careful rather than dismissive: "Reinforcement learning can be part of the solution, but it's not the complete answer. It's not sufficient." Third, self-supervised learning. You tell the machine nothing at all. "And this is what has enabled the recent progress in natural language understanding and chatbots." The strange part is that you stop asking for a task: "You don't train the system to accomplish any particular task. You just train it to basically capture the structure..." The method is deliberate sabotage: "You take a piece of text, you corrupt it in some way, by for example removing some words, and then you train a big neural net to predict the words that are missing." And one narrow version of that trick runs every chatbot on Earth: "A special case of this is that you take a piece of text and the last word in that text is not visible, and so you train the system to predict the last word in that text — and this is the way large language models are trained on." So why did the third one win? Supervised learning needs a human. Reinforcement learning needs a crash. Self-supervised learning needs neither — because the missing word and the correct answer are the same thing. The data grades itself.

Big Brain AI

49,293 görüntüleme • 2 ay önce

Today, Dream Labs is out of stealth. In February 2026, while we were at NVIDIA, Scott Reed youliang Yunhao Ge and I started calling a new kind of robot model a World Action Model, one that learns how the world moves from video, then uses that to decide how to act. Today, we’re releasing Dream-1 through a public API that anyone in the world can use. When you rent a car you’ve never driven, you don’t learn to drive all over again. Within a few minutes, it feels like your own. Robots should work the same way: a good model should be able to pick up a new robot and get to work, without being retrained from scratch. That’s what we’ve spent the last 6 months on, and it’s what Dream-1 does. It controls bimanual robots with parallel grippers it has never seen before, zero-shot, on unseen tasks in unseen environments. Generalization has been my North Star in robotics: first by distilling video world models into VLAs (DreamGen), then by learning to act from those world models directly (DreamZero), and now by transferring across robots (Dream-1). I’ve come to believe it’s the only direction that lasts. Pushing a narrow model from 90% to 99.9% on a single task is real work, but much of it gets washed away once a model that generalizes arrives. Since the start of the company we’ve stayed single-minded about generalization.That focus is what got Dream-1 to control robots it was never trained on, on tasks it was never shown. And it’s what makes a public API possible at all: a model that only does one task on one robot can’t serve millions of users. Dream-1 is still a research beta, and it will make mistakes. But it already drives cars it’s never seen. Generalization wins in the end, and the end is closer than it looks. Here’s Dream-1 taking the wheel of a robot it’s never seen. Full details 👇
0:27

Today, Dream Labs is out of stealth. In February 2026, while we were at NVIDIA, Scott Reed youliang Yunhao Ge and I started calling a new kind of robot model a World Action Model, one that learns how the world moves from video, then uses that to decide how to act. Today, we’re releasing Dream-1 through a public API that anyone in the world can use. When you rent a car you’ve never driven, you don’t learn to drive all over again. Within a few minutes, it feels like your own. Robots should work the same way: a good model should be able to pick up a new robot and get to work, without being retrained from scratch. That’s what we’ve spent the last 6 months on, and it’s what Dream-1 does. It controls bimanual robots with parallel grippers it has never seen before, zero-shot, on unseen tasks in unseen environments. Generalization has been my North Star in robotics: first by distilling video world models into VLAs (DreamGen), then by learning to act from those world models directly (DreamZero), and now by transferring across robots (Dream-1). I’ve come to believe it’s the only direction that lasts. Pushing a narrow model from 90% to 99.9% on a single task is real work, but much of it gets washed away once a model that generalizes arrives. Since the start of the company we’ve stayed single-minded about generalization.That focus is what got Dream-1 to control robots it was never trained on, on tasks it was never shown. And it’s what makes a public API possible at all: a model that only does one task on one robot can’t serve millions of users. Dream-1 is still a research beta, and it will make mistakes. But it already drives cars it’s never seen. Generalization wins in the end, and the end is closer than it looks. Here’s Dream-1 taking the wheel of a robot it’s never seen. Full details 👇

Joel Jang

35,051 görüntüleme • 13 gün önce

This scene in the movie had all the emotions you can think of. But this should have been the final moment for them if it wasn't a movie. IRL, it is obvious he'll pick the daughter above him, and she will always be the second option. I don't think it's going to work that well. She should have gone with Osaze.
4:39

This scene in the movie had all the emotions you can think of. But this should have been the final moment for them if it wasn't a movie. IRL, it is obvious he'll pick the daughter above him, and she will always be the second option. I don't think it's going to work that well. She should have gone with Osaze.

Ìléríolúwa

232,131 görüntüleme • 1 ay önce