Atlas | Partners in Parkour

I can say I'm much impressed. At the same time I haven't watched Boston Dynamics videos for last couple of years. When I checked latest Atlas demo from 2021 it look like sci-fi to me: ▶ this video

We said the same thing about movement in robots -- and today we have: ▶ this video Just ridiculous how far things have come, it seems fake it's so far along.

So they run 5 different experiments to test the hypothesis, and they were nothing like what I imagined. For example, in one study, they divide participants into two groups, have one group watch link (that highlights the high socio-emotional capabilities of a robot), while the other watches ▶ this video (that highlights the low socio-emotional capabilities of a robot) They are then asked if they agree or disagree with a (presumably hypothetical?) company's proposal to reduce employees' welfare, such as replacing a meal with a shake. Two groups showed a different preference. This makes me think about that old question of whether you thank LLM or not. That is treating LLMs more like humans, so if what this paper found holds, maybe that'd nudge our brain subtly toward dehumanizing other real humans!? That's so counter intuitive...

> If anyone is going to make self driving work, the odds are stacked in Tesla's favor. How disconnected from reality are you to make this statement? link Waymo self-driving has been open to general public since June 2024 and Telsa still doesn't have full self driving technology and yet here you are claiming odds are stacked in Tesla's favor. > I feel similarly excited for Tesla when they demonstrate robots with incredible dexterity. Did you never hear about Boston dynamics? ▶ this video This is 3 years old video and Tesla's robots don't have as much dexterity as the robot from the video. And unlike Tesla's robot, Boston Dynamics robot is fully automated.

Two years ago, the Boston Dynamics Atlas could do mid-air somersaults - that's better than most 10-year-olds. ▶ this video

Not yet [1]. Total symbiosis of corporations and government—to the point of making Mussolini blush—is the ultimate goal. Sure, that sounds paranoid, but look at how far we've come on that spectrum in the last 30 years. Proliferation of technology (due to its enabling of surveillance and weakening of the state) all but guarantees that future. [1] ▶ this video

May we finally have walking humanoids robots in our lifetime? I think people are already getting ideas [1] and I am looking forward to everything that follows. From a safe distance There are already walking robots but they are very expensive prototypes and look nothing like humans. [2] Given how many things ChatGPT confidently gets wrong I am not sure how that would play out in a robot. [1] - link [video][movie trailer] [2] - ▶ this video [video][1 min]

That Tim Urban piece is great. It's also an interesting time capsule in terms of which AI problems were and were not considered hard in 2015 (when the post was written). From the post: > Build a computer that can multiply two ten-digit numbers in a split second—incredibly easy. Build one that can look at a dog and answer whether it’s a dog or a cat—spectacularly difficult. Make AI that can beat any human in chess? Done. Make one that can read a paragraph from a six-year-old’s picture book and not just recognize the words but understand the meaning of them? Google is currently spending billions of dollars trying to do it. Hard things—like calculus, financial market strategy, and language translation—are mind-numbingly easy for a computer, while easy things—like vision, motion, movement, and perception—are insanely hard for it. The children's picture book problem is solved; those billions of dollars were well-spent after all. (See, e.g., DeepMind's recent Flamingo model [1].) We can do whatever we want in vision, more or less [2]. Motion and movement might be the least developed area, but it's still made major progress; we have robotic parkour [3] and physical Rubik's cube solvers [4], and we can tell a robot to follow simple domestic instructions [5]. And Perceiver (again from DeepMind [6]) took a big chunk out of the perception problem. Getting a computer to carry on a conversation [7], let alone draw art on par with human professionals [8], weren't even mentioned as examples

The robot I am most personally impressed with is Agility Robotics Digit. It is a real human like form factor (small human not huge!). If I recall correctly it is an energy efficient walking system too. link This long continuous shot of walking from 2019 is nice too. They have improved their software a lot but the hardware was already in good shape. link Meanwhile Boston Dynamics is doing quite well. ▶ this video Best path to success is Tesla buys Agility Robotics, revs the design for Digit to integrate the Tesla AI system, and spins up a manufacturing line. They could focus immediately on the high level AI problem rather than worrying about design. Could slap their AI system on digit while mechanical teams rev the new design so they get moving quickly. Disclosure: I do not work for any of these companies or any related company. I am making a farming robot (see my profile).

I agree with the neighbor comment, 2015 was a long time ago. See the state of the art today: Agility Robotics Digit: link Boston Dynamics Atlas: ▶ this video Neither of those robots use a static walk. That is obvious enough in the above Boston Dynamics video, but Agility Robotics also just released a little video explaining their (very robust) dynamic walking system: link It seems clear the basic hardware to do humanoid walking is technologically achievable today. Personally I would think Tesla would be better suited by copying or buying Agility Robotics as Digit is a smaller robot and it looks cheaper than Atlas too. Better for working in a warehouse alongside people as shown in the top video link in this post, and they seem a smaller company than BD and so cheaper to buy. Musk is very happy to buy firms that produce tech his companies need. As far as the software problem, it is a big one to be sure. I don't think they have all the pieces to the puzzle, but I do think Tesla's Dojo AI system has promise for training a machine like a humanoid. Their system is designed to ingest experience and produce neural nets which can navigate in the real world. They may be willing to take a bet, despite the clear risks, on using this system for humanoids. If they buy Agility Robotics, rev the design to add their AI system (which is extremely powerful AI hardware with their own custom processor), they could be in production on prototypes within a year. This all has the disclaimer that succe

Reusable rockets weren't practical either. But someone was maverick enough to attempt it anyways. [1] You provide valuable information but I just hate how you succumb to easy thinking, an appeal to tradition [2]: "well it hadn't been done then so it can't possibly be done now." It's a form of anachronistic gatekeeping. It doesn't belong in the age we currently reside in. We have totally different computational abilities today, powerful beyond reason. Our hardware is smaller and better performing. We can perform digital simulations billions of times before we even execute one physical experiment. I've seen the same attitude with age-old mathematics problems and it makes even less sense in that domain. The ability to perform computational experiments that our predecessors couldn't, changes the viability of almost every unsolved problem. Boston Dynamics would have never created Atlas or Spot if they clung to common tropes about robots. And now, due to battery technology, the once tethered robots are dexterously mobile. [2] [1] link [2] link [3] ▶ this video

This is the video with the blog post: ▶ this video To me, I like to see when they stutter or slightly-fail, since that to indicates the dynamics and real-time part of the thing, that it is not a 100% scripted thing (although the blog post mentions the engineers fine-tuning the celebratory arm-pump, to what extent is that scripted then?)