August 26, 2026 • 27 mins
We discuss the World Humanoid Olympics in China, where humanoid robots competed in human-style events and set new performance marks. We also cover autonomy, reward-based training, and the comparison between China and the U.S. in robotics and AI.
------
🔒 Check Out Our Sponsor: LEDGER AGENT STACK 🔒
https://developers.ledger.com/docs/ai-tools/overview/?utm_source=Audio&utm_medium=Podcasts&utm_campaign=Limitless
------
🌌 LIMITLESS HQ ⬇️
NEWSLETTER: https://limitlessft.substack.com/
FOLLOW ON X: https://x.com/LimitlessFT
SPOTIFY: https://open.spotify.com/show/5oV29YUL8AzzwXkxEXlRMQ
APPLE: https://podcasts.apple.com/us/podcast/limitless-podcast/id1813210890
RSS FEED: https://limitlessft.substack.com/
------
TIMESTAMPS
0:00 Robot Olympics
1:45 Record-Breaking
8:07 How Robots Learned
9:30 China’s AI Advantage
13:17 Tiangong Takes Lead
16:03 Subsidies and Scale
20:11 Tesla vs China
24:57 Real Benchmark
26:29 Closing
------
RESOURCES
Josh: https://x.com/JoshKale
Ejaaz: https://x.com/cryptopunk7213
------
Not financial or tax advice. See our investment disclosures here:
https://www.bankless.com/disclosures
Josh works with Anthropic as a contractor. All views expressed are his own and do not represent Anthropic, its leadership, or its affiliates. Nothing in this episode is investment advice.
Listen
Watch
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Josh:
26 years ago, a human athlete ran the 1500-meter race in 3 minutes and 26 seconds. (00:00):
undefined
Josh: And ever since then, the record had never been beaten. (00:07):
undefined
Josh: We've had seven Olympics, the latest Nike running shoes, and no one has been able to beat it. (00:10):
undefined
Josh: But over the weekend, a humanoid robot out of China crushed it in 2 minutes (00:14):
undefined
Josh: and 21 seconds. This was the World Humanoid Olympics that was hosted in China over five days. (00:19):
undefined
Josh: And these robots are crushing human-based Olympic events. (00:26):
undefined
Josh: They ran a faster 100 meter than Usain Bolt. They jumped higher than the Olympic record. (00:30):
undefined
Josh: And they broke 1500 meter and the 400 meter races. (00:35):
undefined
Josh: We're talking about 2000 robots from over 16 countries that competed in 55 Olympic (00:38):
undefined
Josh: events. These are the same events that humans compete every four years. (00:44):
undefined
Josh: And the contrast between last year's Robot Olympics and this year's is vastly (00:48):
undefined
Josh: different. And we're going to get into all of that on today's episode. (00:53):
undefined
Ejaaz: This is a visual episode where like, if you're not watching, (00:56):
undefined
Ejaaz: you're probably going to want to watch these videos because they're insane. (00:58):
undefined
Ejaaz: This is also a funny episode because one year ago, we actually filmed the first (01:01):
undefined
Ejaaz: annual Robot Olympics. We are here now for our second annual pretty much any event on the show. (01:05):
undefined
Josh: Yeah, we're on the zone. To talk about the show. (01:09):
undefined
Ejaaz: I mean, dude, just look at the visual of the sound screen. This looks like it's (01:11):
undefined
Ejaaz: AI generated, but it is not AI generated. What we're seeing run across the screen (01:13):
undefined
Ejaaz: is an army of robots that are perfectly in sync. (01:18):
undefined
Ejaaz: For this event, the scale is huge. I think the total amount of robots competing (01:21):
undefined
Ejaaz: is up 300% year over year. (01:26):
undefined
Ejaaz: And the amount of teams is up like 200% year over year. (01:29):
undefined
Ejaaz: And the most interesting statistic is 96% of all the robots that competed in (01:32):
undefined
Ejaaz: this Olympics are all Chinese made. (01:36):
undefined
Ejaaz: This is the China event. And it's kind of sad because I was looking at humanoid (01:39):
undefined
Ejaaz: robots from the US and I was like, oh boy, we might just be cooked. (01:43):
undefined
Ejaaz: So for the records i'm just gonna just list off a couple that are are kind of (01:46):
undefined
Ejaaz: unbelievable and then we'll get into the videos of how they actually did it (01:50):
undefined
Ejaaz: the first is the 100 meter gold that is like the flagship race that everyone kind of (01:52):
undefined
Ejaaz: that denominates in. And that is the one where Usain Bolt famously is the human (01:58):
undefined
Ejaaz: world record. He ran it in 9.58 seconds. (02:02):
undefined
Ejaaz: The robot ran it in 9.39 seconds. And oh my God, look at this video. It's terrifying. (02:04):
undefined
Ejaaz: Could you imagine a robot running at you that fast? (02:11):
undefined
Josh: No. And also look how tall these things are. I could have sworn that these robots (02:13):
undefined
Josh: are meant to be like waist height. (02:18):
undefined
Josh: They were certainly like last year when they were competing. (02:20):
undefined
Josh: This thing is like six foot three to six foot five. These things are huge. (02:22):
undefined
Josh: Oh you see the crash pad over here as well where they just kind of like. (02:26):
undefined
Ejaaz: Slam i find it funny how yeah they crash and they just die like they just turn (02:29):
undefined
Ejaaz: off and then they have to get dragged off in carts they're coming (02:32):
undefined
Josh: Because one and done it's nuts. (02:34):
undefined
Ejaaz: The next one that was kind of crazy is the standing high jump which um (02:37):
undefined
Ejaaz: was above the human record they cleared 2.8 meter 2.88 meters almost nine meters (02:41):
undefined
Ejaaz: the 1500 meter race the tiangang ultra is what they call it was two minutes (02:47):
undefined
Ejaaz: and 21 seconds against a human world record of (02:53):
undefined
Ejaaz: three minutes and 26 seconds, which was untouched since 1998. (02:56):
undefined
Ejaaz: And the robot, most interestingly, these aren't human controlled. (03:02):
undefined
Ejaaz: They ran fully autonomously with zero human intervention. (03:05):
undefined
Ejaaz: So I think all of these things, as we go through the videos from this event, (03:09):
undefined
Ejaaz: yeah, here's the robot running 14.5 miles per hour and just running into the barricade. (03:12):
undefined
Ejaaz: It's the first time that feels scary. And I remember we did this episode a year (03:17):
undefined
Ejaaz: ago and I was like, okay, robots are kind of a joke. Like they don't actually (03:20):
undefined
Ejaaz: look like they are doing anything effective. They're not running that fast. (03:23):
undefined
Ejaaz: They're kind of falling apart and collapsing. (03:26):
undefined
Ejaaz: Similar things were happening this year, but we're seeing some serious upgrades (03:28):
undefined
Ejaaz: year over year. And if we continue this trajectory, even one year in the future, (03:32):
undefined
Ejaaz: that leaves us with a pretty unnerving reality with how capable these robots are gonna be. (03:35):
undefined
Josh: Yeah, I mean, I remember last year's Olympics. And just so that you guys have (03:40):
undefined
Josh: an understanding in how advanced these robots have become in a very short period (03:45):
undefined
Josh: of time, specifically in China. (03:49):
undefined
Josh: Last year, the 100-meter time, the winning robot ran it in 21.5 seconds. (03:51):
undefined
Josh: That's like more than double the world record that Usain Bolt held. (03:57):
undefined
Josh: And now in less than a year, because it has been less than a year, (04:01):
undefined
Josh: they crushed Usain Bolt's record. So what's that going to be next year when (04:04):
undefined
Josh: these robots' arms aren't flailing in a crazy way and they're way more dynamically (04:08):
undefined
Josh: efficient, they're going to crush it even more. (04:12):
undefined
Josh: And so I think the realization that struck to me is this is, (04:14):
undefined
Josh: to your point, no longer a joke. (04:18):
undefined
Josh: The Chinese are very advanced in hardware. We were talking about this before (04:20):
undefined
Josh: we started recording, but hardware and manufacturing has been their kind of (04:24):
undefined
Josh: mecca over the last couple of decades at this point. (04:27):
undefined
Josh: And they have completely beaten the US. When I compare, when I'm looking at (04:30):
undefined
Josh: the figure robots in the US or any kind of robots, Amazon delivery robots. (04:34):
undefined
Josh: I'm just like, China is way more advanced. And for humanoids specifically, (04:38):
undefined
Josh: because this is the other thing, right? It's like, why would these events kind (04:42):
undefined
Josh: of like matter to us at all? (04:46):
undefined
Josh: It's because they're human looking robots competing in human based events. (04:48):
undefined
Josh: Like look, and we're not just talking about, you know, track and field events. (04:52):
undefined
Josh: We're talking about like football. We're talking about like tennis. (04:55):
undefined
Josh: Like I think in this one, there's a crazy penalty. Josh, I want to show you (04:59):
undefined
Josh: this. Hang on one second. (05:02):
undefined
Josh: So a yellow card on the robot and again if you're not watching this I urge you (05:03):
undefined
Josh: to and so he lines up for a penno. (05:07):
undefined
Ejaaz: Drains it absolutely (05:10):
undefined
Josh: Drains it so the point is these robots are no longer toys and we saw the famous (05:12):
undefined
Josh: chinese robot lab unitree ipo last week and their stock is up 650 and that wasn't (05:18):
undefined
Josh: like a pump a dump it's still up around four to five hundred percent (05:24):
undefined
Josh: so there's obviously a demand for these robots these robots are very much at (05:27):
undefined
Josh: the research phase until this event which tied in with a very important conference (05:30):
undefined
Josh: i think it was like the world robotics conference or something like that, (05:35):
undefined
Josh: where the pitch was essentially, these robots are now physically able to do (05:38):
undefined
Josh: a lot of complex things that humans are able to do. (05:41):
undefined
Josh: And you can now apply this to various different industries. And the sports and (05:44):
undefined
Josh: athletics feed is just one example of that. (05:48):
undefined
Ejaaz: There's one thing that I want to highlight that I found extremely interesting (05:50):
undefined
Ejaaz: from this, because one of the new rules that happened this year is that if it's (05:53):
undefined
Ejaaz: a flat event, meaning like if the robot is on a flat surface, (05:56):
undefined
Ejaaz: it must be fully autonomous. (06:00):
undefined
Ejaaz: And this is true for the relay races. This is true for anything on these flat surfaces. (06:02):
undefined
Ejaaz: And what we noticed this year was a property that a lot of people laughed at, (06:05):
undefined
Ejaaz: which is this robot, you guys know you saw it, running with its arms, (06:10):
undefined
Ejaaz: waving, flailing crazy. And this is the best robot. Do we have a video that (06:13):
undefined
Ejaaz: we can show them? Yes, this. (06:16):
undefined
Ejaaz: So look at that robot. Look at how they're running with their arms flailing side to side. (06:18):
undefined
Ejaaz: And I thought this was funny and goofy until I realized that this was actually (06:22):
undefined
Ejaaz: an emergent property of machine learning. (06:26):
undefined
Ejaaz: And this was not trained into the robot this was understood and emerged from (06:29):
undefined
Ejaaz: what yielded the best results and it turns out that like the the robots are (06:35):
undefined
Ejaaz: getting smart enough where they can kind of (06:39):
undefined
Ejaaz: Self-recursively loop themselves and come to terms with the idea that this is (06:41):
undefined
Ejaaz: the optimal strategy and when i look at this i'm laughing because this is kind (06:45):
undefined
Ejaaz: of what evolution developed as a tail (06:49):
undefined
Ejaaz: for a lot of animals that were running around like a lot of these large animals (06:51):
undefined
Ejaaz: that run fast they need a tail as a counterbalance they don't fall over (06:55):
undefined
Ejaaz: these humanoid robots now are running with their arms flailing back and forth (06:58):
undefined
Ejaaz: as a mechanism for balancing. And I wonder, I wonder if this is some sort of (07:02):
undefined
Ejaaz: Move 37 from Go. If you remember, there was like Google's DeepMind computer. (07:06):
undefined
Ejaaz: It was competing against a human. (07:10):
undefined
Ejaaz: And on Move 37, it did this like crazy, unbelievable move that no one expected. (07:12):
undefined
Ejaaz: No one understood why, but it turned out to be genius because it was thinking so far in advance. (07:16):
undefined
Ejaaz: I wonder what types of emergent properties we'll get and we'll learn about the (07:20):
undefined
Ejaaz: human body through this. (07:23):
undefined
Ejaaz: Like perhaps what if Usain Bolt spent his whole life training to run and training (07:24):
undefined
Ejaaz: to run like this instead of with his arms straight forward like would that turn (07:28):
undefined
Ejaaz: into a superhuman runner that's even faster I don't know but I find this really fascinating that (07:32):
undefined
Ejaaz: humans are kind of or the humanoid robots are kind of relearning human movement (07:36):
undefined
Ejaaz: principles from first principles and the results of that are going to be pretty (07:40):
undefined
Ejaaz: interesting as we continue to watch these evolve like how are they going to (07:44):
undefined
Ejaaz: move in a different way than us if if so (07:47):
undefined
Ejaaz: and this is one of the examples I thought was like super interesting really bizarre and funny (07:49):
undefined
Ejaaz: but on the inside like very technically interesting It's like, (07:53):
undefined
Ejaaz: oh, they created their own tail. That's kind of cool. (07:55):
undefined
Josh: When I look at this, it seems like it's in a perpetual state of falling, (07:58):
undefined
Josh: like it's about to stumble, but it just gets in front of it just at the right time. (08:03):
undefined
Josh: Now, the story behind this is actually interesting, just kind of like unpack it a bit more. (08:07):
undefined
Josh: The makers of this robot in particular that crushed the 400 meter record, (08:11):
undefined
Josh: a human record, is called Tiang Gong. (08:17):
undefined
Josh: And the engineers actually had no idea how to teach this robot to run fast. (08:19):
undefined
Josh: So they started off with the human movement, which is we swing our arms back (08:24):
undefined
Josh: and forth, right? It propels us forward. (08:29):
undefined
Josh: And so they built a robot that did exactly that. And they realized that this (08:31):
undefined
Josh: was incredibly inefficient for their robot. (08:34):
undefined
Josh: Its rotator cuff, its shoulder started getting really heated and it ended up (08:36):
undefined
Josh: exploding. So its arms would literally fall off in the test case when they had (08:41):
undefined
Josh: a robot that would run like a human. (08:45):
undefined
Josh: So they didn't know, they were scratching their heads. They were like, (08:47):
undefined
Josh: how on earth do we figure out how to make this robot run super fast autonomously? (08:49):
undefined
Josh: And so it just gave the robot model that was in its brain the task and said, (08:53):
undefined
Josh: I will give you more rewards if you figure out how to run faster. (08:58):
undefined
Josh: Your goal is to get to the finish line as quick as you can. (09:01):
undefined
Josh: And the robot started doing this. It started swinging its arm from side to side, (09:04):
undefined
Josh: which basically mimicked the swinging your arms back and forth, (09:10):
undefined
Josh: forwards as you're running normally like a human, but it had less resistance (09:14):
undefined
Josh: on the shoulder rotator cuff. (09:18):
undefined
Josh: So it figured it out on itself independently. (09:19):
undefined
Josh: And this might be a more uniform way of running. I can't speak on behalf of (09:22):
undefined
Josh: humans, but certainly for robots, it ran the fastest time it has ever run. (09:26):
undefined
Josh: I saw another stat that was super interesting. 70% of AI tokens that are used (09:30):
undefined
Josh: in China are apparently used for video models specifically. (09:36):
undefined
Josh: And a large chunk of that is for robot models. So when we look at this and we're (09:41):
undefined
Josh: wondering how on earth is China like, you know, usurp the US? (09:46):
undefined
Josh: How have they like become so advanced in such a short amount of time? (09:49):
undefined
Josh: It's because they're spent the majority of their AI tokens and research development (09:52):
undefined
Josh: on video models specifically. (09:56):
undefined
Josh: And we see this with like the text of video models that we've seen outside of (09:57):
undefined
Josh: this that we've talked about on the show on various other episodes (10:00):
undefined
Josh: they are far superior than the u.s based labs and so this makes sense like when (10:03):
undefined
Josh: it comes to the world of world models and robotic brains it just has they have (10:06):
undefined
Josh: a they have us on visual intelligence is what i'm realizing as i'm watching (10:11):
undefined
Josh: more and more of these videos. (10:14):
undefined
Ejaaz: One thing that i did want to show i just sent you a post if you don't mind pulling (10:15):
undefined
Ejaaz: it up yeah because i thought this was so funny too this is what an earlier version of this (10:18):
undefined
Ejaaz: emergence of understanding human movement looked like this was an early google (10:22):
undefined
Ejaaz: game mind project and what they did is they gave this like figure human properties (10:26):
undefined
Ejaaz: but didn't teach it how to walk (10:31):
undefined
Ejaaz: and you can see it slowly learning over time how to move its joints and its (10:33):
undefined
Ejaaz: body wow to figure out how to like actually get from point a to point b using (10:37):
undefined
Ejaaz: a reward system that you just described which is like if you can get to this (10:41):
undefined
Ejaaz: place you will be rewarded figure out how to get there and you see it really yeah (10:44):
undefined
Ejaaz: really like weird because it is like a humanoid looking thing and what we're (10:49):
undefined
Ejaaz: seeing today is not only the evolution of this early technology that we saw at a google deep mind (10:53):
undefined
Ejaaz: but the iterative improvement so much so that it actually has a physical body (10:57):
undefined
Ejaaz: to live within when i think about the next evolution of or the next phase of (11:00):
undefined
Ejaaz: this like general evolution as we get greater intelligence naturally the (11:05):
undefined
Ejaaz: The hope is to embody that intelligence, to put it into these physical objects (11:10):
undefined
Ejaaz: that are going to be moving physical atoms around. This is the next thing. (11:14):
undefined
Ejaaz: It's much more difficult because we have to manufacture things. (11:17):
undefined
Ejaaz: We're not the best at manufacturing things. (11:19):
undefined
Ejaaz: But in the case that we are competing head to head with China on manufacturing (11:20):
undefined
Ejaaz: things and applying AI to it, I see no world in which we stand a chance. (11:24):
undefined
Ejaaz: And that's a kind of scary place to be because like 96% of these robots that (11:29):
undefined
Ejaaz: competed are Chinese. And when I look at American robots, they're different. (11:34):
undefined
Ejaaz: They're different. They're more general purpose. It's much more technically challenging. (11:39):
undefined
Josh: They suck. (11:42):
undefined
Ejaaz: They currently do suck. Like, it's funny. I was scrolling my feed. (11:44):
undefined
Ejaaz: We were talking about this yesterday. (11:47):
undefined
Josh: Even this robot's head is down. Look at this, Josh. Like at the start, look at it. (11:48):
undefined
Ejaaz: Oh, he's just so sad. He watched the Robot Olympics. So I'm scrolling my feed. (11:52):
undefined
Ejaaz: I'm seeing these robots doing like backflips. They're running 20 miles an hour. (11:55):
undefined
Ejaaz: They're like doing high jumps. They're doing these unbelievable things. (11:58):
undefined
Ejaaz: And then Brett Adcock from Figure, who is a humanoid robot maker, (12:01):
undefined
Ejaaz: he's sharing like just robots kind of walking around campus slowly. (12:04):
undefined
Ejaaz: And he shared a video of like a robot slowly walking up a staircase yeah this (12:07):
undefined
Ejaaz: one he's like look guys he climbed like 15 feet of stairs and i'm like man i (12:10):
undefined
Ejaaz: understand these are different robots i understand this is a general purpose (12:16):
undefined
Ejaaz: robot that can do a lot of things (12:19):
undefined
Ejaaz: just on a purely optical basis when you look at the videos from the olympics (12:21):
undefined
Ejaaz: versus when you're looking at the videos from this you're like man we got a long way to go (12:25):
undefined
Ejaaz: and not only that but we have what two maybe three companies here in the u.s (12:29):
undefined
Ejaaz: that are doing it we have figure we have tesla (12:32):
undefined
Ejaaz: maybe there's like boston dynamics china everyone's doing this there is like no (12:34):
undefined
Ejaaz: specific company that is taking the lead everyone's building these unbelievable (12:39):
undefined
Ejaaz: robots assuming that this progress continues next year is going to look freaking (12:42):
undefined
Ejaaz: crazy like i cannot imagine they're probably going to need a bigger arena because (12:46):
undefined
Ejaaz: i'm sure the robots are going to very quickly outgrow this like small confined (12:50):
undefined
Ejaaz: space originally meant for humans (12:52):
undefined
Ejaaz: this was taking place in the olympic stadium that is used for humans uh traditionally (12:54):
undefined
Ejaaz: i think it's like the 200 meter lap race or whatever it is but really like unbelievable (12:59):
undefined
Ejaaz: rate of progress and like kind of a scary (13:03):
undefined
Ejaaz: thing to see knowing that none of those robots belong to us (13:06):
undefined
Josh: So when i'm looking at this i'm thinking okay we know china is obviously very (13:09):
undefined
Josh: advanced compared to the u.s but which companies in particular in china are (13:13):
undefined
Josh: they private? Are they public? (13:17):
undefined
Josh: Like what's so special about them? What are the name brands that we need to (13:18):
undefined
Josh: kind of like bring ourselves to be aware of? (13:21):
undefined
Josh: The winner, the winning brand is this company called Tiang Gong. (13:23):
undefined
Josh: I don't know if I'm pronouncing that correctly, but Tiang Gong. (13:27):
undefined
Josh: And it's basically a startup from Beijing, China. And no one kind of like knew (13:30):
undefined
Josh: about their name very well outside of China, at least. (13:35):
undefined
Josh: I think the name brand was basically Unitary, which placed second across all (13:38):
undefined
Josh: the medals that were won so far. (13:42):
undefined
Josh: Remember, it's still going on as you're listening to this right now, (13:43):
undefined
Josh: you can probably catch the live stream. (13:45):
undefined
Josh: But Tiangong, it's a small startup. (13:47):
undefined
Josh: And they crushed it at the 400 meter, as well as a number of other track events. (13:50):
undefined
Josh: And the reason why they're getting a lot of the glory right now is, (13:54):
undefined
Josh: to your earlier point, any kind of flat surface events had to be completely (13:57):
undefined
Josh: autonomous. So there was no human controller at all. (14:00):
undefined
Josh: So it's a much harder feat to teach or train an AI robot to basically run around (14:03):
undefined
Josh: a track or kind of like maneuver around various different obstacle courses. (14:08):
undefined
Josh: And Tiangong had the best AI model paired with their hardware, (14:12):
undefined
Josh: right? So it was very kind of like pliable. (14:16):
undefined
Josh: Now, when you look at Unitree, Unitree is a name brand that I mentioned earlier, (14:18):
undefined
Josh: that IPO last week, and their stock is up like something like 600%. (14:21):
undefined
Josh: And I think they're valued at like $55 billion, where they initially IPO'd up (14:25):
undefined
Josh: like $2 billion. So you're seeing these crazy multiple returns in China as of (14:30):
undefined
Josh: recent. And it's not just in robotics. It's happening in the semiconductor space (14:34):
undefined
Josh: with memory stuff. I think they had the CXMT IPO. (14:37):
undefined
Josh: China's having a bit of a moment with their AI models and with their robotics (14:40):
undefined
Josh: side of things. and I just want to spend a quick moment kind of describing what (14:44):
undefined
Josh: those conjectures sort of look like. (14:47):
undefined
Josh: On the AI model side of things, they have kept up in open source and now they're (14:49):
undefined
Josh: kind of putting out models that are competing on the frontier. (14:54):
undefined
Josh: There's this model, I don't know if you've seen it, being tested recently, (14:57):
undefined
Josh: Josh, but it's called Ox Alpha and many people think that it's meant to be GLM's (14:59):
undefined
Josh: next model, 5.3, and it's a flash model. (15:04):
undefined
Ejaaz: Stay tuned for that episode once we know. (15:06):
undefined
Josh: It's going to be a banging episode, hopefully, once they officially release it. (15:07):
undefined
Josh: But that is competing at the level of a Fable 5 or a GPT 5.6 Sol. (15:10):
undefined
Josh: And it's a hell of a lot cheaper. So somehow Chinese have been able to keep (15:14):
undefined
Josh: up on the software model side. (15:18):
undefined
Josh: But if you flip over to the hardware side, which it is my strong belief is going (15:20):
undefined
Josh: to be the next kind of phase or frontier of AI, we're talking about world models, (15:24):
undefined
Josh: visual intelligence, all that kind of stuff, we see a flow. (15:28):
undefined
Josh: Where China actually might have the more advanced models and advanced hardware (15:31):
undefined
Josh: and the scale to pump these robots out. Remember, China has, (15:36):
undefined
Josh: I think, something ridiculous, like five times more robots. (15:40):
undefined
Josh: Oh, they produce five times more robots than the US did in all of 2025. (15:44):
undefined
Josh: And that's scaling on an exponential curve where the US just can't compete. (15:49):
undefined
Josh: Right now, it's just like two startups. (15:51):
undefined
Josh: So China's having a bit of a moment and it's something that we need to keep (15:53):
undefined
Josh: an eye on, both from an investment perspective, but also from like protecting (15:56):
undefined
Josh: the frontier if the West wants to stay ahead. of, you know, their main adversaries. (15:59):
undefined
Ejaaz: And it also, it feels like it's private industry versus the government in the (16:04):
undefined
Ejaaz: US versus China. Like this is interesting. (16:08):
undefined
Ejaaz: Yeah. The Beijing government, they poured $20 billion in subsidies into the (16:10):
undefined
Ejaaz: sector over just the last year alone, and reportedly is planning a 1 trillion yuan fund. (16:15):
undefined
Ejaaz: I'm not sure exactly what that translates to, but that is like a tremendous (16:20):
undefined
Ejaaz: amount of money that they're putting towards these companies to push the frontier forward. (16:23):
undefined
Ejaaz: And this is in a way China's advantage (16:27):
undefined
Ejaaz: where they are able to just force progress by any means necessary. (16:30):
undefined
Ejaaz: We've seen this many instances with Frontier AI labs where they will just subsidize (16:34):
undefined
Ejaaz: price or even prior to AI. (16:40):
undefined
Ejaaz: If there was a competing product from China and a competing product from the (16:42):
undefined
Ejaaz: United States, what they would do is the Chinese government would step in. (16:44):
undefined
Ejaaz: They're going to say, hey, we're going to subsidize you. (16:47):
undefined
Ejaaz: We're going to make the cost lower so that you can outcompete everyone else. (16:49):
undefined
Ejaaz: The entire market will go to you and then we'll gradually raise the prices up. (16:53):
undefined
Ejaaz: But they're willing to step in and put their hand on the scale to make meaningful (16:56):
undefined
Ejaaz: change in the progress of a specific division. And it's very clear that robots (17:00):
undefined
Ejaaz: is one of their top priorities. (17:04):
undefined
Ejaaz: And this is interesting. It's happened in the past around manufacturing. (17:07):
undefined
Ejaaz: I know China has this like requirement for a lot of ships where when they're (17:10):
undefined
Ejaaz: made, they must be made to a size in which you could fit a tank on top of them. (17:14):
undefined
Ejaaz: Like you need to be able to roll a tank on top of your boat for it to be classified (17:18):
undefined
Ejaaz: as a specific type of boat that you could then go off and build and use and (17:22):
undefined
Ejaaz: get permits for because they they're thinking in this like almost military sense (17:24):
undefined
Ejaaz: and they have the control over everyone where they can do that. (17:29):
undefined
Ejaaz: So I think it's really interesting to see the dynamics of kind of like private (17:31):
undefined
Ejaaz: government versus public industry at play (17:35):
undefined
Ejaaz: um here in the u.s versus china and so far like capitalism kicking ass but man (17:38):
undefined
Ejaaz: like china with their subsidies and their (17:43):
undefined
Ejaaz: their ability to manufacture things at scale it seems like a legitimate concern (17:45):
undefined
Ejaaz: and sure a lot of this is staged a lot of the software is like pre-coded but (17:50):
undefined
Ejaaz: the robots are real like those things are (17:54):
undefined
Ejaaz: real they're produced at scale they're doing the flips they're running the things (17:57):
undefined
Ejaaz: and they're getting better much faster so if we gauge this on a rate of progress (18:00):
undefined
Ejaaz: which is all we can ever do (18:03):
undefined
Ejaaz: uh year over year i'd say they're up like what is this a 10x i remember last (18:05):
undefined
Ejaaz: year as we were sitting here talking about this it was mostly a joke we were (18:08):
undefined
Ejaaz: just kind of laughing like hot look at that robot like falling sparks flying everywhere (18:11):
undefined
Ejaaz: just like dropping dead and totally blanking out but this year it is a very (18:15):
undefined
Ejaaz: different story and like next year (18:18):
undefined
Ejaaz: Man, you got to imagine it's just going to be even more slightly deranged than this. (18:20):
undefined
Josh: I'm not even sure how much of this year's robots were hard-coded. (18:25):
undefined
Josh: Like when I look at that 400 meter race and that robot kind of developed that (18:30):
undefined
Josh: weird sideways swinging movement by itself, I think like, wow, (18:35):
undefined
Josh: okay, so next year's robots are probably going to come up with completely new (18:38):
undefined
Josh: performant ways that humans (18:41):
undefined
Josh: probably could learn from at the actual human Olympics. (18:43):
undefined
Josh: So I'm curious how that develops. (18:46):
undefined
Josh: But your point around kind of like having a nation state influence and dictate (18:48):
undefined
Josh: how a certain technology progresses, in this case, robotics, (18:53):
undefined
Josh: it has had meaningful impact. (18:56):
undefined
Josh: And listen, like on the West, when I compare it, right, we are arguing to ban (18:58):
undefined
Josh: data centers almost everywhere in every single state in the US, at least, right? (19:03):
undefined
Josh: And that is going to hinder progress in massive ways. (19:07):
undefined
Josh: And listen, some of the reasoning is very valid around electricity concerns. (19:11):
undefined
Josh: But I know that a lot of the AI labs are taking very meaningful steps to subsidize (19:15):
undefined
Josh: that and also pay a much larger tax so that it goes to the local community so (19:20):
undefined
Josh: they can get much more benefits from this kind of stuff. (19:24):
undefined
Josh: There's a lot of pushback. The general feel and vibe in the Western Front, (19:27):
undefined
Josh: or I'll speak for the US, is AI is bad. It consumes a lot of water. (19:32):
undefined
Josh: It is bad for humans. It is taking our jobs. It is replacing us. (19:36):
undefined
Josh: And I'm not saying that's right or wrong. But when you compare that to China, (19:39):
undefined
Josh: which is so AI forward, it's just so robotics forward, they are obviously going to get the leap on us. (19:42):
undefined
Josh: The US government comparatively is like, hmm, should we ban open models? (19:47):
undefined
Josh: Should we ban some of these other frontier AI models? And China, (19:51):
undefined
Josh: meanwhile, is like, we're going to do anything and everything to get ahead of (19:53):
undefined
Josh: this space because we know it's going to be incredibly transformative technology (19:57):
undefined
Josh: and we'll make it safe when we can. (19:59):
undefined
Josh: I don't know what the right path is forwards, but we are going to see a stark (20:01):
undefined
Josh: contrast over the next year is my guess if the two countries don't change their (20:05):
undefined
Josh: kind of ideological views on how this pans out. (20:09):
undefined
Josh: What I would like to see, Josh, next year is Tesla Optimus compete in this. (20:12):
undefined
Josh: I'm sure or I know Elon Musk is looking at this and thinking, (20:18):
undefined
Josh: hmm, I think I have a card up my sleeve and I can do this. (20:22):
undefined
Josh: So I'm hoping that the Chinese Olympics, Robot Olympics will be open to taking (20:27):
undefined
Josh: a Tesla Optimus because I kind of want to see how badass it is compared to it. (20:32):
undefined
Josh: Now, China, granted, has all the data and information that is required to train frontier robotics. (20:35):
undefined
Josh: If you look at any of the hardware components, like the actuators, (20:41):
undefined
Josh: the rotator cuffs, they produce the majority of it. (20:44):
undefined
Josh: I think actually Tesla Optimus sources a bunch of this from China, (20:46):
undefined
Josh: But I want to see the Western front kind of like compete on this because right (20:49):
undefined
Josh: now this feels like the GPT-3 moment Where China had nothing and American had (20:52):
undefined
Josh: all the frontier AI models, but swapped the other way around. (20:57):
undefined
Ejaaz: As you're saying that I'm like man, why is it China versus Elon? (21:01):
undefined
Ejaaz: Why is it not China versus US? Why do we only have Elon and Optimus to put up against him? (21:03):
undefined
Ejaaz: Maybe the figure robots It is also important to note to play both sides of the (21:07):
undefined
Ejaaz: coin here that the United States is (21:11):
undefined
Ejaaz: trying very different strategies than China in the most cases. (21:14):
undefined
Ejaaz: Like Optimus and the figure robots are general purpose humanoid robots that (21:17):
undefined
Ejaaz: are met for very different use cases than the ones that we're watching on screen today. (21:22):
undefined
Ejaaz: The ones on screen today are very narrow purpose. They're built for a singular (21:27):
undefined
Ejaaz: specific thing, whether it be racing, all these things are within a closed loop (21:31):
undefined
Ejaaz: system. So we're seeing pickleball so crazy too. (21:34):
undefined
Ejaaz: How long do you think, or is this tennis? I can't even, it's tennis. (21:37):
undefined
Ejaaz: How long do you think until a robot can like, EJS, I know you play a paddle (21:41):
undefined
Ejaaz: all the time like yes what would you give the over under on like say two years (21:44):
undefined
Ejaaz: before a robot can kick your ass oh (21:48):
undefined
Josh: Well i mean i would probably give it, (21:50):
undefined
Josh: two maybe three years because like look at look at how well okay right now this (21:53):
undefined
Josh: is like the equivalent of me laughing at this right because it looks like there's (21:58):
undefined
Josh: a young kid playing tennis over here but i said that last year (22:01):
undefined
Josh: for all the track of it's and now they're beating the world record so if i assume (22:04):
undefined
Josh: in a year's time this is going to beat like say carlos alcaraz or djokovic then (22:09):
undefined
Josh: yeah actually maybe even a year it could. (22:14):
undefined
Ejaaz: Happen could happen it could happen and the thing with these systems is like (22:17):
undefined
Ejaaz: humanoids that we're trying to build at tesla or figure those need to work for (22:20):
undefined
Ejaaz: the real world those need to be trained on everything because they need to be prepared for (22:25):
undefined
Ejaaz: the unknown variables that occur on a day-to-day basis that (22:29):
undefined
Ejaaz: no computer can really predict what these are doing is they're being trained on very narrow focused (22:32):
undefined
Ejaaz: versions of that so like this is a ping pong trained robot this ping pong robot (22:37):
undefined
Ejaaz: cannot go and play tennis it cannot go run the 100 meter dash and win gold it (22:41):
undefined
Ejaaz: is specific for this one use case and it's very cool to see but it is just a (22:46):
undefined
Ejaaz: totally different strategy and when i think about (22:49):
undefined
Ejaaz: where value is going to be derived from humanoids it's certainly across or from (22:51):
undefined
Ejaaz: just like robots in general it's certainly across the stack but a general purpose (22:55):
undefined
Ejaaz: humanoid who can do useful productive tasks is going to generally be more productive than a (22:58):
undefined
Ejaaz: humanoid robot that can like kick like oh my god did you see (23:04):
undefined
Josh: That do you see that that's great can you rewind that for everyone. (23:07):
undefined
Ejaaz: For those of you listening this is like a full 360 roundhouse kick and kicked the robot's head off. (23:12):
undefined
Josh: This is not AI. This is completely real. (23:17):
undefined
Ejaaz: That's so sick, man. This is basically the- You're like walking out of a bar (23:19):
undefined
Ejaaz: or something and a robot starts picking a fight with you. (23:24):
undefined
Josh: And it starts hitting the emotes. The Fortnite emotes. (23:26):
undefined
Ejaaz: Look at this. It's so insane. I do appreciate that they're having fun with it. (23:29):
undefined
Ejaaz: But yeah, like that's it. That's what's happening. Like that is the robot Olympics (23:33):
undefined
Ejaaz: summed up is like, it's scary. Like robots don't suck anymore. Robots are- (23:37):
undefined
Ejaaz: Clearly the future, they're clearly being produced at scale, (23:41):
undefined
Ejaaz: and they're clearly finding these emergent properties of human beings that we (23:44):
undefined
Ejaaz: have not even considered to try yet. (23:47):
undefined
Ejaaz: And like, this is so cool. If anything, just a really fun visual event. (23:51):
undefined
Ejaaz: I really enjoyed scrolling this weekend, seeing all the videos from it and just (23:54):
undefined
Ejaaz: kind of watching in awe at how much progress they've made just in the last 12 months. (23:57):
undefined
Ejaaz: It's pretty sick. And what's also sick is agents, because if you are not fully (24:00):
undefined
Ejaaz: operating humanoid robots, chances are you're running agentic robots. (24:05):
undefined
Ejaaz: And security is a fairly big concern when you're working with agents. (24:08):
undefined
Ejaaz: And we have a solution for that, which comes in the form of Ledger, (24:11):
undefined
Ejaaz: our friend and sponsor of the show. (24:13):
undefined
Ejaaz: If you are building with agents, you are probably worried about security. (24:15):
undefined
Ejaaz: Ledger has an agent stack that fixes this. It does it in three steps. (24:18):
undefined
Ejaaz: The agent proposes, the humans approve, the Ledger signer enforces. (24:21):
undefined
Ejaaz: And it comes with this entire stack of software that's open source fully available (24:25):
undefined
Ejaaz: and helpful for you to navigate your security concerns as it relates to engaging (24:29):
undefined
Ejaaz: with agents it works with cloud code codex cursor basically anywhere that you (24:33):
undefined
Ejaaz: integrate and work with ai and it's fully open source and linked in the description (24:37):
undefined
Ejaaz: below so if you are in need of security for agents go check out ledger and thank (24:41):
undefined
Ejaaz: you to ledger for sponsoring this part of the episode and with that (24:44):
undefined
Ejaaz: i think we have described the robot olympics (24:47):
undefined
Josh: The great part about just seeing these robots perform like people might think (24:51):
undefined
Josh: this is like novel and like kind of stupid. (24:55):
undefined
Josh: It is sort of a real physical benchmark. Like you can't game this benchmark. (24:58):
undefined
Josh: This isn't like an AI model that kind of like fakes its scores and kind of like (25:03):
undefined
Josh: scores really well against some (25:07):
undefined
Josh: kind of like arbitrary exam where you can kind of see the answer sheet. (25:08):
undefined
Josh: This is if a robot can jump high enough, you're going to be able to see it. (25:11):
undefined
Josh: If it can kick a free throw or penalty and hit a Sue afterwards, (25:14):
undefined
Josh: hell yeah, it can do that as well. (25:18):
undefined
Josh: Like the point is, I like that with robots in particular, there is no kind of (25:19):
undefined
Josh: like hidden in plain sight type of thing. (25:25):
undefined
Josh: If a robot is good, you'll be able to see it performing as well as it can. (25:27):
undefined
Josh: And I think China right now, if I was to summarize, is taking the number one (25:30):
undefined
Josh: spot from the US and it's the US position to respond to. (25:35):
undefined
Josh: And two, it's broken a number of the human Olympic records. Again, (25:40):
undefined
Josh: just to recap, 1500 meter record is now owned by a humanoid robot. (25:43):
undefined
Josh: The 100 meter record, Usain Bolt, sorry, goodbye, mate. Like it's owned by Tiangong, right? 400 meter. (25:48):
undefined
Josh: Sorry, it's owned by a robot that runs like this with its arms swinging side (25:54):
undefined
Josh: to side. It looks stupid, but they're taking over, right? (25:58):
undefined
Josh: Now, imagine if you retrofit a robot that is as agile as this to your manufacturing (26:01):
undefined
Josh: plant or to a factory or to any kind of manual labor job. (26:07):
undefined
Josh: If it can do this, it can probably do some simple kind of movement, (26:11):
undefined
Josh: such as packaging and stuff like that. (26:13):
undefined
Josh: And my expectation is this is just the chaser. This is the advertising to say, (26:15):
undefined
Josh: hey, don't you think we're cool? (26:20):
undefined
Josh: Now let's have a serious business chat around what these robots can do for you. (26:22):
undefined
Josh: And so, yeah, it's fun and novel, but I think we're going to see some real progress (26:26):
undefined
Josh: in robotics going forward. (26:29):
undefined
Josh: And I am pumped to see it. So if you've listened to this episode, (26:30):
undefined
Josh: that is the robotics download. There's anything and everything you need to know. (26:33):
undefined
Josh: In the last year, China is winning very effectively. If you're listening to (26:37):
undefined
Josh: this on YouTube, Spotify, or wherever you're listening to, please give us a (26:40):
undefined
Josh: thumbs up. Leave us a comment. (26:43):
undefined
Josh: It helps us out massively. We read every single comment. There are a lot of (26:46):
undefined
Josh: newbies, Josh. I think it's like 1,500 over the last couple of weeks. (26:49):
undefined
Ejaaz: Welcome to the party. (26:53):
undefined
Josh: Welcome to the party. It is lovely to have you here. We post around four times a week. (26:54):
undefined
Josh: We love talking about AI, robotics, and any kind of frontier tech. (26:59):
undefined
Josh: If there's anything that we've missed over the last couple of weeks that you (27:02):
undefined
Josh: think we should include, let us know. Reach out to us on our personal profiles. (27:05):
undefined
Josh: Josh, any final concluding thoughts? (27:09):
undefined
Ejaaz: Yeah, don't forget to share with a friend who you want to terrify. (27:11):
undefined
Ejaaz: You know who you could share with? (27:14):
undefined
Ejaaz: There's a lot of people who still use AI as a Google search, (27:16):
undefined
Ejaaz: and they're not really sure about what's going on. (27:21):
undefined
Ejaaz: Just send him a video from the humanoid olympics be like listen this is where (27:22):
undefined
Ejaaz: we're at it's a lot farther along than you think and maybe he'll give him a little scare (27:25):
undefined
Ejaaz: before the end of the day but yeah it's it's nuts it's not so we're going to (27:30):
undefined
Ejaaz: be following along with that as well as all the other fun and exciting topics (27:33):
undefined
Ejaaz: on the frontier of technology of ai and thank you just so much for being here (27:37):
undefined
Ejaaz: with us for the journey so with that we will see you guys on the next one (27:41):
undefined
undefined
Josh: And ever since then, the record had never been beaten. (00:07):
undefined
Josh: We've had seven Olympics, the latest Nike running shoes, and no one has been able to beat it. (00:10):
undefined
Josh: But over the weekend, a humanoid robot out of China crushed it in 2 minutes (00:14):
undefined
Josh: and 21 seconds. This was the World Humanoid Olympics that was hosted in China over five days. (00:19):
undefined
Josh: And these robots are crushing human-based Olympic events. (00:26):
undefined
Josh: They ran a faster 100 meter than Usain Bolt. They jumped higher than the Olympic record. (00:30):
undefined
Josh: And they broke 1500 meter and the 400 meter races. (00:35):
undefined
Josh: We're talking about 2000 robots from over 16 countries that competed in 55 Olympic (00:38):
undefined
Josh: events. These are the same events that humans compete every four years. (00:44):
undefined
Josh: And the contrast between last year's Robot Olympics and this year's is vastly (00:48):
undefined
Josh: different. And we're going to get into all of that on today's episode. (00:53):
undefined
Ejaaz: This is a visual episode where like, if you're not watching, (00:56):
undefined
Ejaaz: you're probably going to want to watch these videos because they're insane. (00:58):
undefined
Ejaaz: This is also a funny episode because one year ago, we actually filmed the first (01:01):
undefined
Ejaaz: annual Robot Olympics. We are here now for our second annual pretty much any event on the show. (01:05):
undefined
Josh: Yeah, we're on the zone. To talk about the show. (01:09):
undefined
Ejaaz: I mean, dude, just look at the visual of the sound screen. This looks like it's (01:11):
undefined
Ejaaz: AI generated, but it is not AI generated. What we're seeing run across the screen (01:13):
undefined
Ejaaz: is an army of robots that are perfectly in sync. (01:18):
undefined
Ejaaz: For this event, the scale is huge. I think the total amount of robots competing (01:21):
undefined
Ejaaz: is up 300% year over year. (01:26):
undefined
Ejaaz: And the amount of teams is up like 200% year over year. (01:29):
undefined
Ejaaz: And the most interesting statistic is 96% of all the robots that competed in (01:32):
undefined
Ejaaz: this Olympics are all Chinese made. (01:36):
undefined
Ejaaz: This is the China event. And it's kind of sad because I was looking at humanoid (01:39):
undefined
Ejaaz: robots from the US and I was like, oh boy, we might just be cooked. (01:43):
undefined
Ejaaz: So for the records i'm just gonna just list off a couple that are are kind of (01:46):
undefined
Ejaaz: unbelievable and then we'll get into the videos of how they actually did it (01:50):
undefined
Ejaaz: the first is the 100 meter gold that is like the flagship race that everyone kind of (01:52):
undefined
Ejaaz: that denominates in. And that is the one where Usain Bolt famously is the human (01:58):
undefined
Ejaaz: world record. He ran it in 9.58 seconds. (02:02):
undefined
Ejaaz: The robot ran it in 9.39 seconds. And oh my God, look at this video. It's terrifying. (02:04):
undefined
Ejaaz: Could you imagine a robot running at you that fast? (02:11):
undefined
Josh: No. And also look how tall these things are. I could have sworn that these robots (02:13):
undefined
Josh: are meant to be like waist height. (02:18):
undefined
Josh: They were certainly like last year when they were competing. (02:20):
undefined
Josh: This thing is like six foot three to six foot five. These things are huge. (02:22):
undefined
Josh: Oh you see the crash pad over here as well where they just kind of like. (02:26):
undefined
Ejaaz: Slam i find it funny how yeah they crash and they just die like they just turn (02:29):
undefined
Ejaaz: off and then they have to get dragged off in carts they're coming (02:32):
undefined
Josh: Because one and done it's nuts. (02:34):
undefined
Ejaaz: The next one that was kind of crazy is the standing high jump which um (02:37):
undefined
Ejaaz: was above the human record they cleared 2.8 meter 2.88 meters almost nine meters (02:41):
undefined
Ejaaz: the 1500 meter race the tiangang ultra is what they call it was two minutes (02:47):
undefined
Ejaaz: and 21 seconds against a human world record of (02:53):
undefined
Ejaaz: three minutes and 26 seconds, which was untouched since 1998. (02:56):
undefined
Ejaaz: And the robot, most interestingly, these aren't human controlled. (03:02):
undefined
Ejaaz: They ran fully autonomously with zero human intervention. (03:05):
undefined
Ejaaz: So I think all of these things, as we go through the videos from this event, (03:09):
undefined
Ejaaz: yeah, here's the robot running 14.5 miles per hour and just running into the barricade. (03:12):
undefined
Ejaaz: It's the first time that feels scary. And I remember we did this episode a year (03:17):
undefined
Ejaaz: ago and I was like, okay, robots are kind of a joke. Like they don't actually (03:20):
undefined
Ejaaz: look like they are doing anything effective. They're not running that fast. (03:23):
undefined
Ejaaz: They're kind of falling apart and collapsing. (03:26):
undefined
Ejaaz: Similar things were happening this year, but we're seeing some serious upgrades (03:28):
undefined
Ejaaz: year over year. And if we continue this trajectory, even one year in the future, (03:32):
undefined
Ejaaz: that leaves us with a pretty unnerving reality with how capable these robots are gonna be. (03:35):
undefined
Josh: Yeah, I mean, I remember last year's Olympics. And just so that you guys have (03:40):
undefined
Josh: an understanding in how advanced these robots have become in a very short period (03:45):
undefined
Josh: of time, specifically in China. (03:49):
undefined
Josh: Last year, the 100-meter time, the winning robot ran it in 21.5 seconds. (03:51):
undefined
Josh: That's like more than double the world record that Usain Bolt held. (03:57):
undefined
Josh: And now in less than a year, because it has been less than a year, (04:01):
undefined
Josh: they crushed Usain Bolt's record. So what's that going to be next year when (04:04):
undefined
Josh: these robots' arms aren't flailing in a crazy way and they're way more dynamically (04:08):
undefined
Josh: efficient, they're going to crush it even more. (04:12):
undefined
Josh: And so I think the realization that struck to me is this is, (04:14):
undefined
Josh: to your point, no longer a joke. (04:18):
undefined
Josh: The Chinese are very advanced in hardware. We were talking about this before (04:20):
undefined
Josh: we started recording, but hardware and manufacturing has been their kind of (04:24):
undefined
Josh: mecca over the last couple of decades at this point. (04:27):
undefined
Josh: And they have completely beaten the US. When I compare, when I'm looking at (04:30):
undefined
Josh: the figure robots in the US or any kind of robots, Amazon delivery robots. (04:34):
undefined
Josh: I'm just like, China is way more advanced. And for humanoids specifically, (04:38):
undefined
Josh: because this is the other thing, right? It's like, why would these events kind (04:42):
undefined
Josh: of like matter to us at all? (04:46):
undefined
Josh: It's because they're human looking robots competing in human based events. (04:48):
undefined
Josh: Like look, and we're not just talking about, you know, track and field events. (04:52):
undefined
Josh: We're talking about like football. We're talking about like tennis. (04:55):
undefined
Josh: Like I think in this one, there's a crazy penalty. Josh, I want to show you (04:59):
undefined
Josh: this. Hang on one second. (05:02):
undefined
Josh: So a yellow card on the robot and again if you're not watching this I urge you (05:03):
undefined
Josh: to and so he lines up for a penno. (05:07):
undefined
Ejaaz: Drains it absolutely (05:10):
undefined
Josh: Drains it so the point is these robots are no longer toys and we saw the famous (05:12):
undefined
Josh: chinese robot lab unitree ipo last week and their stock is up 650 and that wasn't (05:18):
undefined
Josh: like a pump a dump it's still up around four to five hundred percent (05:24):
undefined
Josh: so there's obviously a demand for these robots these robots are very much at (05:27):
undefined
Josh: the research phase until this event which tied in with a very important conference (05:30):
undefined
Josh: i think it was like the world robotics conference or something like that, (05:35):
undefined
Josh: where the pitch was essentially, these robots are now physically able to do (05:38):
undefined
Josh: a lot of complex things that humans are able to do. (05:41):
undefined
Josh: And you can now apply this to various different industries. And the sports and (05:44):
undefined
Josh: athletics feed is just one example of that. (05:48):
undefined
Ejaaz: There's one thing that I want to highlight that I found extremely interesting (05:50):
undefined
Ejaaz: from this, because one of the new rules that happened this year is that if it's (05:53):
undefined
Ejaaz: a flat event, meaning like if the robot is on a flat surface, (05:56):
undefined
Ejaaz: it must be fully autonomous. (06:00):
undefined
Ejaaz: And this is true for the relay races. This is true for anything on these flat surfaces. (06:02):
undefined
Ejaaz: And what we noticed this year was a property that a lot of people laughed at, (06:05):
undefined
Ejaaz: which is this robot, you guys know you saw it, running with its arms, (06:10):
undefined
Ejaaz: waving, flailing crazy. And this is the best robot. Do we have a video that (06:13):
undefined
Ejaaz: we can show them? Yes, this. (06:16):
undefined
Ejaaz: So look at that robot. Look at how they're running with their arms flailing side to side. (06:18):
undefined
Ejaaz: And I thought this was funny and goofy until I realized that this was actually (06:22):
undefined
Ejaaz: an emergent property of machine learning. (06:26):
undefined
Ejaaz: And this was not trained into the robot this was understood and emerged from (06:29):
undefined
Ejaaz: what yielded the best results and it turns out that like the the robots are (06:35):
undefined
Ejaaz: getting smart enough where they can kind of (06:39):
undefined
Ejaaz: Self-recursively loop themselves and come to terms with the idea that this is (06:41):
undefined
Ejaaz: the optimal strategy and when i look at this i'm laughing because this is kind (06:45):
undefined
Ejaaz: of what evolution developed as a tail (06:49):
undefined
Ejaaz: for a lot of animals that were running around like a lot of these large animals (06:51):
undefined
Ejaaz: that run fast they need a tail as a counterbalance they don't fall over (06:55):
undefined
Ejaaz: these humanoid robots now are running with their arms flailing back and forth (06:58):
undefined
Ejaaz: as a mechanism for balancing. And I wonder, I wonder if this is some sort of (07:02):
undefined
Ejaaz: Move 37 from Go. If you remember, there was like Google's DeepMind computer. (07:06):
undefined
Ejaaz: It was competing against a human. (07:10):
undefined
Ejaaz: And on Move 37, it did this like crazy, unbelievable move that no one expected. (07:12):
undefined
Ejaaz: No one understood why, but it turned out to be genius because it was thinking so far in advance. (07:16):
undefined
Ejaaz: I wonder what types of emergent properties we'll get and we'll learn about the (07:20):
undefined
Ejaaz: human body through this. (07:23):
undefined
Ejaaz: Like perhaps what if Usain Bolt spent his whole life training to run and training (07:24):
undefined
Ejaaz: to run like this instead of with his arms straight forward like would that turn (07:28):
undefined
Ejaaz: into a superhuman runner that's even faster I don't know but I find this really fascinating that (07:32):
undefined
Ejaaz: humans are kind of or the humanoid robots are kind of relearning human movement (07:36):
undefined
Ejaaz: principles from first principles and the results of that are going to be pretty (07:40):
undefined
Ejaaz: interesting as we continue to watch these evolve like how are they going to (07:44):
undefined
Ejaaz: move in a different way than us if if so (07:47):
undefined
Ejaaz: and this is one of the examples I thought was like super interesting really bizarre and funny (07:49):
undefined
Ejaaz: but on the inside like very technically interesting It's like, (07:53):
undefined
Ejaaz: oh, they created their own tail. That's kind of cool. (07:55):
undefined
Josh: When I look at this, it seems like it's in a perpetual state of falling, (07:58):
undefined
Josh: like it's about to stumble, but it just gets in front of it just at the right time. (08:03):
undefined
Josh: Now, the story behind this is actually interesting, just kind of like unpack it a bit more. (08:07):
undefined
Josh: The makers of this robot in particular that crushed the 400 meter record, (08:11):
undefined
Josh: a human record, is called Tiang Gong. (08:17):
undefined
Josh: And the engineers actually had no idea how to teach this robot to run fast. (08:19):
undefined
Josh: So they started off with the human movement, which is we swing our arms back (08:24):
undefined
Josh: and forth, right? It propels us forward. (08:29):
undefined
Josh: And so they built a robot that did exactly that. And they realized that this (08:31):
undefined
Josh: was incredibly inefficient for their robot. (08:34):
undefined
Josh: Its rotator cuff, its shoulder started getting really heated and it ended up (08:36):
undefined
Josh: exploding. So its arms would literally fall off in the test case when they had (08:41):
undefined
Josh: a robot that would run like a human. (08:45):
undefined
Josh: So they didn't know, they were scratching their heads. They were like, (08:47):
undefined
Josh: how on earth do we figure out how to make this robot run super fast autonomously? (08:49):
undefined
Josh: And so it just gave the robot model that was in its brain the task and said, (08:53):
undefined
Josh: I will give you more rewards if you figure out how to run faster. (08:58):
undefined
Josh: Your goal is to get to the finish line as quick as you can. (09:01):
undefined
Josh: And the robot started doing this. It started swinging its arm from side to side, (09:04):
undefined
Josh: which basically mimicked the swinging your arms back and forth, (09:10):
undefined
Josh: forwards as you're running normally like a human, but it had less resistance (09:14):
undefined
Josh: on the shoulder rotator cuff. (09:18):
undefined
Josh: So it figured it out on itself independently. (09:19):
undefined
Josh: And this might be a more uniform way of running. I can't speak on behalf of (09:22):
undefined
Josh: humans, but certainly for robots, it ran the fastest time it has ever run. (09:26):
undefined
Josh: I saw another stat that was super interesting. 70% of AI tokens that are used (09:30):
undefined
Josh: in China are apparently used for video models specifically. (09:36):
undefined
Josh: And a large chunk of that is for robot models. So when we look at this and we're (09:41):
undefined
Josh: wondering how on earth is China like, you know, usurp the US? (09:46):
undefined
Josh: How have they like become so advanced in such a short amount of time? (09:49):
undefined
Josh: It's because they're spent the majority of their AI tokens and research development (09:52):
undefined
Josh: on video models specifically. (09:56):
undefined
Josh: And we see this with like the text of video models that we've seen outside of (09:57):
undefined
Josh: this that we've talked about on the show on various other episodes (10:00):
undefined
Josh: they are far superior than the u.s based labs and so this makes sense like when (10:03):
undefined
Josh: it comes to the world of world models and robotic brains it just has they have (10:06):
undefined
Josh: a they have us on visual intelligence is what i'm realizing as i'm watching (10:11):
undefined
Josh: more and more of these videos. (10:14):
undefined
Ejaaz: One thing that i did want to show i just sent you a post if you don't mind pulling (10:15):
undefined
Ejaaz: it up yeah because i thought this was so funny too this is what an earlier version of this (10:18):
undefined
Ejaaz: emergence of understanding human movement looked like this was an early google (10:22):
undefined
Ejaaz: game mind project and what they did is they gave this like figure human properties (10:26):
undefined
Ejaaz: but didn't teach it how to walk (10:31):
undefined
Ejaaz: and you can see it slowly learning over time how to move its joints and its (10:33):
undefined
Ejaaz: body wow to figure out how to like actually get from point a to point b using (10:37):
undefined
Ejaaz: a reward system that you just described which is like if you can get to this (10:41):
undefined
Ejaaz: place you will be rewarded figure out how to get there and you see it really yeah (10:44):
undefined
Ejaaz: really like weird because it is like a humanoid looking thing and what we're (10:49):
undefined
Ejaaz: seeing today is not only the evolution of this early technology that we saw at a google deep mind (10:53):
undefined
Ejaaz: but the iterative improvement so much so that it actually has a physical body (10:57):
undefined
Ejaaz: to live within when i think about the next evolution of or the next phase of (11:00):
undefined
Ejaaz: this like general evolution as we get greater intelligence naturally the (11:05):
undefined
Ejaaz: The hope is to embody that intelligence, to put it into these physical objects (11:10):
undefined
Ejaaz: that are going to be moving physical atoms around. This is the next thing. (11:14):
undefined
Ejaaz: It's much more difficult because we have to manufacture things. (11:17):
undefined
Ejaaz: We're not the best at manufacturing things. (11:19):
undefined
Ejaaz: But in the case that we are competing head to head with China on manufacturing (11:20):
undefined
Ejaaz: things and applying AI to it, I see no world in which we stand a chance. (11:24):
undefined
Ejaaz: And that's a kind of scary place to be because like 96% of these robots that (11:29):
undefined
Ejaaz: competed are Chinese. And when I look at American robots, they're different. (11:34):
undefined
Ejaaz: They're different. They're more general purpose. It's much more technically challenging. (11:39):
undefined
Josh: They suck. (11:42):
undefined
Ejaaz: They currently do suck. Like, it's funny. I was scrolling my feed. (11:44):
undefined
Ejaaz: We were talking about this yesterday. (11:47):
undefined
Josh: Even this robot's head is down. Look at this, Josh. Like at the start, look at it. (11:48):
undefined
Ejaaz: Oh, he's just so sad. He watched the Robot Olympics. So I'm scrolling my feed. (11:52):
undefined
Ejaaz: I'm seeing these robots doing like backflips. They're running 20 miles an hour. (11:55):
undefined
Ejaaz: They're like doing high jumps. They're doing these unbelievable things. (11:58):
undefined
Ejaaz: And then Brett Adcock from Figure, who is a humanoid robot maker, (12:01):
undefined
Ejaaz: he's sharing like just robots kind of walking around campus slowly. (12:04):
undefined
Ejaaz: And he shared a video of like a robot slowly walking up a staircase yeah this (12:07):
undefined
Ejaaz: one he's like look guys he climbed like 15 feet of stairs and i'm like man i (12:10):
undefined
Ejaaz: understand these are different robots i understand this is a general purpose (12:16):
undefined
Ejaaz: robot that can do a lot of things (12:19):
undefined
Ejaaz: just on a purely optical basis when you look at the videos from the olympics (12:21):
undefined
Ejaaz: versus when you're looking at the videos from this you're like man we got a long way to go (12:25):
undefined
Ejaaz: and not only that but we have what two maybe three companies here in the u.s (12:29):
undefined
Ejaaz: that are doing it we have figure we have tesla (12:32):
undefined
Ejaaz: maybe there's like boston dynamics china everyone's doing this there is like no (12:34):
undefined
Ejaaz: specific company that is taking the lead everyone's building these unbelievable (12:39):
undefined
Ejaaz: robots assuming that this progress continues next year is going to look freaking (12:42):
undefined
Ejaaz: crazy like i cannot imagine they're probably going to need a bigger arena because (12:46):
undefined
Ejaaz: i'm sure the robots are going to very quickly outgrow this like small confined (12:50):
undefined
Ejaaz: space originally meant for humans (12:52):
undefined
Ejaaz: this was taking place in the olympic stadium that is used for humans uh traditionally (12:54):
undefined
Ejaaz: i think it's like the 200 meter lap race or whatever it is but really like unbelievable (12:59):
undefined
Ejaaz: rate of progress and like kind of a scary (13:03):
undefined
Ejaaz: thing to see knowing that none of those robots belong to us (13:06):
undefined
Josh: So when i'm looking at this i'm thinking okay we know china is obviously very (13:09):
undefined
Josh: advanced compared to the u.s but which companies in particular in china are (13:13):
undefined
Josh: they private? Are they public? (13:17):
undefined
Josh: Like what's so special about them? What are the name brands that we need to (13:18):
undefined
Josh: kind of like bring ourselves to be aware of? (13:21):
undefined
Josh: The winner, the winning brand is this company called Tiang Gong. (13:23):
undefined
Josh: I don't know if I'm pronouncing that correctly, but Tiang Gong. (13:27):
undefined
Josh: And it's basically a startup from Beijing, China. And no one kind of like knew (13:30):
undefined
Josh: about their name very well outside of China, at least. (13:35):
undefined
Josh: I think the name brand was basically Unitary, which placed second across all (13:38):
undefined
Josh: the medals that were won so far. (13:42):
undefined
Josh: Remember, it's still going on as you're listening to this right now, (13:43):
undefined
Josh: you can probably catch the live stream. (13:45):
undefined
Josh: But Tiangong, it's a small startup. (13:47):
undefined
Josh: And they crushed it at the 400 meter, as well as a number of other track events. (13:50):
undefined
Josh: And the reason why they're getting a lot of the glory right now is, (13:54):
undefined
Josh: to your earlier point, any kind of flat surface events had to be completely (13:57):
undefined
Josh: autonomous. So there was no human controller at all. (14:00):
undefined
Josh: So it's a much harder feat to teach or train an AI robot to basically run around (14:03):
undefined
Josh: a track or kind of like maneuver around various different obstacle courses. (14:08):
undefined
Josh: And Tiangong had the best AI model paired with their hardware, (14:12):
undefined
Josh: right? So it was very kind of like pliable. (14:16):
undefined
Josh: Now, when you look at Unitree, Unitree is a name brand that I mentioned earlier, (14:18):
undefined
Josh: that IPO last week, and their stock is up like something like 600%. (14:21):
undefined
Josh: And I think they're valued at like $55 billion, where they initially IPO'd up (14:25):
undefined
Josh: like $2 billion. So you're seeing these crazy multiple returns in China as of (14:30):
undefined
Josh: recent. And it's not just in robotics. It's happening in the semiconductor space (14:34):
undefined
Josh: with memory stuff. I think they had the CXMT IPO. (14:37):
undefined
Josh: China's having a bit of a moment with their AI models and with their robotics (14:40):
undefined
Josh: side of things. and I just want to spend a quick moment kind of describing what (14:44):
undefined
Josh: those conjectures sort of look like. (14:47):
undefined
Josh: On the AI model side of things, they have kept up in open source and now they're (14:49):
undefined
Josh: kind of putting out models that are competing on the frontier. (14:54):
undefined
Josh: There's this model, I don't know if you've seen it, being tested recently, (14:57):
undefined
Josh: Josh, but it's called Ox Alpha and many people think that it's meant to be GLM's (14:59):
undefined
Josh: next model, 5.3, and it's a flash model. (15:04):
undefined
Ejaaz: Stay tuned for that episode once we know. (15:06):
undefined
Josh: It's going to be a banging episode, hopefully, once they officially release it. (15:07):
undefined
Josh: But that is competing at the level of a Fable 5 or a GPT 5.6 Sol. (15:10):
undefined
Josh: And it's a hell of a lot cheaper. So somehow Chinese have been able to keep (15:14):
undefined
Josh: up on the software model side. (15:18):
undefined
Josh: But if you flip over to the hardware side, which it is my strong belief is going (15:20):
undefined
Josh: to be the next kind of phase or frontier of AI, we're talking about world models, (15:24):
undefined
Josh: visual intelligence, all that kind of stuff, we see a flow. (15:28):
undefined
Josh: Where China actually might have the more advanced models and advanced hardware (15:31):
undefined
Josh: and the scale to pump these robots out. Remember, China has, (15:36):
undefined
Josh: I think, something ridiculous, like five times more robots. (15:40):
undefined
Josh: Oh, they produce five times more robots than the US did in all of 2025. (15:44):
undefined
Josh: And that's scaling on an exponential curve where the US just can't compete. (15:49):
undefined
Josh: Right now, it's just like two startups. (15:51):
undefined
Josh: So China's having a bit of a moment and it's something that we need to keep (15:53):
undefined
Josh: an eye on, both from an investment perspective, but also from like protecting (15:56):
undefined
Josh: the frontier if the West wants to stay ahead. of, you know, their main adversaries. (15:59):
undefined
Ejaaz: And it also, it feels like it's private industry versus the government in the (16:04):
undefined
Ejaaz: US versus China. Like this is interesting. (16:08):
undefined
Ejaaz: Yeah. The Beijing government, they poured $20 billion in subsidies into the (16:10):
undefined
Ejaaz: sector over just the last year alone, and reportedly is planning a 1 trillion yuan fund. (16:15):
undefined
Ejaaz: I'm not sure exactly what that translates to, but that is like a tremendous (16:20):
undefined
Ejaaz: amount of money that they're putting towards these companies to push the frontier forward. (16:23):
undefined
Ejaaz: And this is in a way China's advantage (16:27):
undefined
Ejaaz: where they are able to just force progress by any means necessary. (16:30):
undefined
Ejaaz: We've seen this many instances with Frontier AI labs where they will just subsidize (16:34):
undefined
Ejaaz: price or even prior to AI. (16:40):
undefined
Ejaaz: If there was a competing product from China and a competing product from the (16:42):
undefined
Ejaaz: United States, what they would do is the Chinese government would step in. (16:44):
undefined
Ejaaz: They're going to say, hey, we're going to subsidize you. (16:47):
undefined
Ejaaz: We're going to make the cost lower so that you can outcompete everyone else. (16:49):
undefined
Ejaaz: The entire market will go to you and then we'll gradually raise the prices up. (16:53):
undefined
Ejaaz: But they're willing to step in and put their hand on the scale to make meaningful (16:56):
undefined
Ejaaz: change in the progress of a specific division. And it's very clear that robots (17:00):
undefined
Ejaaz: is one of their top priorities. (17:04):
undefined
Ejaaz: And this is interesting. It's happened in the past around manufacturing. (17:07):
undefined
Ejaaz: I know China has this like requirement for a lot of ships where when they're (17:10):
undefined
Ejaaz: made, they must be made to a size in which you could fit a tank on top of them. (17:14):
undefined
Ejaaz: Like you need to be able to roll a tank on top of your boat for it to be classified (17:18):
undefined
Ejaaz: as a specific type of boat that you could then go off and build and use and (17:22):
undefined
Ejaaz: get permits for because they they're thinking in this like almost military sense (17:24):
undefined
Ejaaz: and they have the control over everyone where they can do that. (17:29):
undefined
Ejaaz: So I think it's really interesting to see the dynamics of kind of like private (17:31):
undefined
Ejaaz: government versus public industry at play (17:35):
undefined
Ejaaz: um here in the u.s versus china and so far like capitalism kicking ass but man (17:38):
undefined
Ejaaz: like china with their subsidies and their (17:43):
undefined
Ejaaz: their ability to manufacture things at scale it seems like a legitimate concern (17:45):
undefined
Ejaaz: and sure a lot of this is staged a lot of the software is like pre-coded but (17:50):
undefined
Ejaaz: the robots are real like those things are (17:54):
undefined
Ejaaz: real they're produced at scale they're doing the flips they're running the things (17:57):
undefined
Ejaaz: and they're getting better much faster so if we gauge this on a rate of progress (18:00):
undefined
Ejaaz: which is all we can ever do (18:03):
undefined
Ejaaz: uh year over year i'd say they're up like what is this a 10x i remember last (18:05):
undefined
Ejaaz: year as we were sitting here talking about this it was mostly a joke we were (18:08):
undefined
Ejaaz: just kind of laughing like hot look at that robot like falling sparks flying everywhere (18:11):
undefined
Ejaaz: just like dropping dead and totally blanking out but this year it is a very (18:15):
undefined
Ejaaz: different story and like next year (18:18):
undefined
Ejaaz: Man, you got to imagine it's just going to be even more slightly deranged than this. (18:20):
undefined
Josh: I'm not even sure how much of this year's robots were hard-coded. (18:25):
undefined
Josh: Like when I look at that 400 meter race and that robot kind of developed that (18:30):
undefined
Josh: weird sideways swinging movement by itself, I think like, wow, (18:35):
undefined
Josh: okay, so next year's robots are probably going to come up with completely new (18:38):
undefined
Josh: performant ways that humans (18:41):
undefined
Josh: probably could learn from at the actual human Olympics. (18:43):
undefined
Josh: So I'm curious how that develops. (18:46):
undefined
Josh: But your point around kind of like having a nation state influence and dictate (18:48):
undefined
Josh: how a certain technology progresses, in this case, robotics, (18:53):
undefined
Josh: it has had meaningful impact. (18:56):
undefined
Josh: And listen, like on the West, when I compare it, right, we are arguing to ban (18:58):
undefined
Josh: data centers almost everywhere in every single state in the US, at least, right? (19:03):
undefined
Josh: And that is going to hinder progress in massive ways. (19:07):
undefined
Josh: And listen, some of the reasoning is very valid around electricity concerns. (19:11):
undefined
Josh: But I know that a lot of the AI labs are taking very meaningful steps to subsidize (19:15):
undefined
Josh: that and also pay a much larger tax so that it goes to the local community so (19:20):
undefined
Josh: they can get much more benefits from this kind of stuff. (19:24):
undefined
Josh: There's a lot of pushback. The general feel and vibe in the Western Front, (19:27):
undefined
Josh: or I'll speak for the US, is AI is bad. It consumes a lot of water. (19:32):
undefined
Josh: It is bad for humans. It is taking our jobs. It is replacing us. (19:36):
undefined
Josh: And I'm not saying that's right or wrong. But when you compare that to China, (19:39):
undefined
Josh: which is so AI forward, it's just so robotics forward, they are obviously going to get the leap on us. (19:42):
undefined
Josh: The US government comparatively is like, hmm, should we ban open models? (19:47):
undefined
Josh: Should we ban some of these other frontier AI models? And China, (19:51):
undefined
Josh: meanwhile, is like, we're going to do anything and everything to get ahead of (19:53):
undefined
Josh: this space because we know it's going to be incredibly transformative technology (19:57):
undefined
Josh: and we'll make it safe when we can. (19:59):
undefined
Josh: I don't know what the right path is forwards, but we are going to see a stark (20:01):
undefined
Josh: contrast over the next year is my guess if the two countries don't change their (20:05):
undefined
Josh: kind of ideological views on how this pans out. (20:09):
undefined
Josh: What I would like to see, Josh, next year is Tesla Optimus compete in this. (20:12):
undefined
Josh: I'm sure or I know Elon Musk is looking at this and thinking, (20:18):
undefined
Josh: hmm, I think I have a card up my sleeve and I can do this. (20:22):
undefined
Josh: So I'm hoping that the Chinese Olympics, Robot Olympics will be open to taking (20:27):
undefined
Josh: a Tesla Optimus because I kind of want to see how badass it is compared to it. (20:32):
undefined
Josh: Now, China, granted, has all the data and information that is required to train frontier robotics. (20:35):
undefined
Josh: If you look at any of the hardware components, like the actuators, (20:41):
undefined
Josh: the rotator cuffs, they produce the majority of it. (20:44):
undefined
Josh: I think actually Tesla Optimus sources a bunch of this from China, (20:46):
undefined
Josh: But I want to see the Western front kind of like compete on this because right (20:49):
undefined
Josh: now this feels like the GPT-3 moment Where China had nothing and American had (20:52):
undefined
Josh: all the frontier AI models, but swapped the other way around. (20:57):
undefined
Ejaaz: As you're saying that I'm like man, why is it China versus Elon? (21:01):
undefined
Ejaaz: Why is it not China versus US? Why do we only have Elon and Optimus to put up against him? (21:03):
undefined
Ejaaz: Maybe the figure robots It is also important to note to play both sides of the (21:07):
undefined
Ejaaz: coin here that the United States is (21:11):
undefined
Ejaaz: trying very different strategies than China in the most cases. (21:14):
undefined
Ejaaz: Like Optimus and the figure robots are general purpose humanoid robots that (21:17):
undefined
Ejaaz: are met for very different use cases than the ones that we're watching on screen today. (21:22):
undefined
Ejaaz: The ones on screen today are very narrow purpose. They're built for a singular (21:27):
undefined
Ejaaz: specific thing, whether it be racing, all these things are within a closed loop (21:31):
undefined
Ejaaz: system. So we're seeing pickleball so crazy too. (21:34):
undefined
Ejaaz: How long do you think, or is this tennis? I can't even, it's tennis. (21:37):
undefined
Ejaaz: How long do you think until a robot can like, EJS, I know you play a paddle (21:41):
undefined
Ejaaz: all the time like yes what would you give the over under on like say two years (21:44):
undefined
Ejaaz: before a robot can kick your ass oh (21:48):
undefined
Josh: Well i mean i would probably give it, (21:50):
undefined
Josh: two maybe three years because like look at look at how well okay right now this (21:53):
undefined
Josh: is like the equivalent of me laughing at this right because it looks like there's (21:58):
undefined
Josh: a young kid playing tennis over here but i said that last year (22:01):
undefined
Josh: for all the track of it's and now they're beating the world record so if i assume (22:04):
undefined
Josh: in a year's time this is going to beat like say carlos alcaraz or djokovic then (22:09):
undefined
Josh: yeah actually maybe even a year it could. (22:14):
undefined
Ejaaz: Happen could happen it could happen and the thing with these systems is like (22:17):
undefined
Ejaaz: humanoids that we're trying to build at tesla or figure those need to work for (22:20):
undefined
Ejaaz: the real world those need to be trained on everything because they need to be prepared for (22:25):
undefined
Ejaaz: the unknown variables that occur on a day-to-day basis that (22:29):
undefined
Ejaaz: no computer can really predict what these are doing is they're being trained on very narrow focused (22:32):
undefined
Ejaaz: versions of that so like this is a ping pong trained robot this ping pong robot (22:37):
undefined
Ejaaz: cannot go and play tennis it cannot go run the 100 meter dash and win gold it (22:41):
undefined
Ejaaz: is specific for this one use case and it's very cool to see but it is just a (22:46):
undefined
Ejaaz: totally different strategy and when i think about (22:49):
undefined
Ejaaz: where value is going to be derived from humanoids it's certainly across or from (22:51):
undefined
Ejaaz: just like robots in general it's certainly across the stack but a general purpose (22:55):
undefined
Ejaaz: humanoid who can do useful productive tasks is going to generally be more productive than a (22:58):
undefined
Ejaaz: humanoid robot that can like kick like oh my god did you see (23:04):
undefined
Josh: That do you see that that's great can you rewind that for everyone. (23:07):
undefined
Ejaaz: For those of you listening this is like a full 360 roundhouse kick and kicked the robot's head off. (23:12):
undefined
Josh: This is not AI. This is completely real. (23:17):
undefined
Ejaaz: That's so sick, man. This is basically the- You're like walking out of a bar (23:19):
undefined
Ejaaz: or something and a robot starts picking a fight with you. (23:24):
undefined
Josh: And it starts hitting the emotes. The Fortnite emotes. (23:26):
undefined
Ejaaz: Look at this. It's so insane. I do appreciate that they're having fun with it. (23:29):
undefined
Ejaaz: But yeah, like that's it. That's what's happening. Like that is the robot Olympics (23:33):
undefined
Ejaaz: summed up is like, it's scary. Like robots don't suck anymore. Robots are- (23:37):
undefined
Ejaaz: Clearly the future, they're clearly being produced at scale, (23:41):
undefined
Ejaaz: and they're clearly finding these emergent properties of human beings that we (23:44):
undefined
Ejaaz: have not even considered to try yet. (23:47):
undefined
Ejaaz: And like, this is so cool. If anything, just a really fun visual event. (23:51):
undefined
Ejaaz: I really enjoyed scrolling this weekend, seeing all the videos from it and just (23:54):
undefined
Ejaaz: kind of watching in awe at how much progress they've made just in the last 12 months. (23:57):
undefined
Ejaaz: It's pretty sick. And what's also sick is agents, because if you are not fully (24:00):
undefined
Ejaaz: operating humanoid robots, chances are you're running agentic robots. (24:05):
undefined
Ejaaz: And security is a fairly big concern when you're working with agents. (24:08):
undefined
Ejaaz: And we have a solution for that, which comes in the form of Ledger, (24:11):
undefined
Ejaaz: our friend and sponsor of the show. (24:13):
undefined
Ejaaz: If you are building with agents, you are probably worried about security. (24:15):
undefined
Ejaaz: Ledger has an agent stack that fixes this. It does it in three steps. (24:18):
undefined
Ejaaz: The agent proposes, the humans approve, the Ledger signer enforces. (24:21):
undefined
Ejaaz: And it comes with this entire stack of software that's open source fully available (24:25):
undefined
Ejaaz: and helpful for you to navigate your security concerns as it relates to engaging (24:29):
undefined
Ejaaz: with agents it works with cloud code codex cursor basically anywhere that you (24:33):
undefined
Ejaaz: integrate and work with ai and it's fully open source and linked in the description (24:37):
undefined
Ejaaz: below so if you are in need of security for agents go check out ledger and thank (24:41):
undefined
Ejaaz: you to ledger for sponsoring this part of the episode and with that (24:44):
undefined
Ejaaz: i think we have described the robot olympics (24:47):
undefined
Josh: The great part about just seeing these robots perform like people might think (24:51):
undefined
Josh: this is like novel and like kind of stupid. (24:55):
undefined
Josh: It is sort of a real physical benchmark. Like you can't game this benchmark. (24:58):
undefined
Josh: This isn't like an AI model that kind of like fakes its scores and kind of like (25:03):
undefined
Josh: scores really well against some (25:07):
undefined
Josh: kind of like arbitrary exam where you can kind of see the answer sheet. (25:08):
undefined
Josh: This is if a robot can jump high enough, you're going to be able to see it. (25:11):
undefined
Josh: If it can kick a free throw or penalty and hit a Sue afterwards, (25:14):
undefined
Josh: hell yeah, it can do that as well. (25:18):
undefined
Josh: Like the point is, I like that with robots in particular, there is no kind of (25:19):
undefined
Josh: like hidden in plain sight type of thing. (25:25):
undefined
Josh: If a robot is good, you'll be able to see it performing as well as it can. (25:27):
undefined
Josh: And I think China right now, if I was to summarize, is taking the number one (25:30):
undefined
Josh: spot from the US and it's the US position to respond to. (25:35):
undefined
Josh: And two, it's broken a number of the human Olympic records. Again, (25:40):
undefined
Josh: just to recap, 1500 meter record is now owned by a humanoid robot. (25:43):
undefined
Josh: The 100 meter record, Usain Bolt, sorry, goodbye, mate. Like it's owned by Tiangong, right? 400 meter. (25:48):
undefined
Josh: Sorry, it's owned by a robot that runs like this with its arms swinging side (25:54):
undefined
Josh: to side. It looks stupid, but they're taking over, right? (25:58):
undefined
Josh: Now, imagine if you retrofit a robot that is as agile as this to your manufacturing (26:01):
undefined
Josh: plant or to a factory or to any kind of manual labor job. (26:07):
undefined
Josh: If it can do this, it can probably do some simple kind of movement, (26:11):
undefined
Josh: such as packaging and stuff like that. (26:13):
undefined
Josh: And my expectation is this is just the chaser. This is the advertising to say, (26:15):
undefined
Josh: hey, don't you think we're cool? (26:20):
undefined
Josh: Now let's have a serious business chat around what these robots can do for you. (26:22):
undefined
Josh: And so, yeah, it's fun and novel, but I think we're going to see some real progress (26:26):
undefined
Josh: in robotics going forward. (26:29):
undefined
Josh: And I am pumped to see it. So if you've listened to this episode, (26:30):
undefined
Josh: that is the robotics download. There's anything and everything you need to know. (26:33):
undefined
Josh: In the last year, China is winning very effectively. If you're listening to (26:37):
undefined
Josh: this on YouTube, Spotify, or wherever you're listening to, please give us a (26:40):
undefined
Josh: thumbs up. Leave us a comment. (26:43):
undefined
Josh: It helps us out massively. We read every single comment. There are a lot of (26:46):
undefined
Josh: newbies, Josh. I think it's like 1,500 over the last couple of weeks. (26:49):
undefined
Ejaaz: Welcome to the party. (26:53):
undefined
Josh: Welcome to the party. It is lovely to have you here. We post around four times a week. (26:54):
undefined
Josh: We love talking about AI, robotics, and any kind of frontier tech. (26:59):
undefined
Josh: If there's anything that we've missed over the last couple of weeks that you (27:02):
undefined
Josh: think we should include, let us know. Reach out to us on our personal profiles. (27:05):
undefined
Josh: Josh, any final concluding thoughts? (27:09):
undefined
Ejaaz: Yeah, don't forget to share with a friend who you want to terrify. (27:11):
undefined
Ejaaz: You know who you could share with? (27:14):
undefined
Ejaaz: There's a lot of people who still use AI as a Google search, (27:16):
undefined
Ejaaz: and they're not really sure about what's going on. (27:21):
undefined
Ejaaz: Just send him a video from the humanoid olympics be like listen this is where (27:22):
undefined
Ejaaz: we're at it's a lot farther along than you think and maybe he'll give him a little scare (27:25):
undefined
Ejaaz: before the end of the day but yeah it's it's nuts it's not so we're going to (27:30):
undefined
Ejaaz: be following along with that as well as all the other fun and exciting topics (27:33):
undefined
Ejaaz: on the frontier of technology of ai and thank you just so much for being here (27:37):
undefined
Ejaaz: with us for the journey so with that we will see you guys on the next one (27:41):
undefined
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
The Breakfast Club
The World's Most Dangerous Morning Show, The Breakfast Club, With DJ Envy, Jess Hilarious, And Charlamagne Tha God!