Sabitlenmiş Tweet
What I’ve been working on 🙂
Bringing a creature like this to life using reinforcement learning has always been a dream of mine. It’s incredible to see it becoming real.
English
Chris Nota
144 posts
Chris Nota
@chris_nota_rl
Founding Researcher @ Familiar Machines & Magic. Reinforcement learning, creature robotics, and artificial life.
Katılım Mayıs 2026
287 Takip Edilen88 Takipçiler
I envision a world filled with artificial creatures. Some will be companions. Some will have jobs. Some will just be part of the scenery.
This will happen because creatures bring fascination and joy to many people. They enrich the world we live in.
English
Attension is all you need is old, the new one is "Enhanced attention is all you need"
Giovanni Monea@giomonea
Transformers use the same hidden state for two competing roles: maintaining state and predicting the next token. 🔀 Splitting these roles into two separate streams outperforms standard Transformers. Introducing the State-Prediction Separation (SPS) Transformer. 🧵
English
Artificial life needs more than just LLM agents!
hardmaru@hardmaru
Language models and coding agents are great, but there is more to life, and more to AI, than just LLM agents.
English
Last month, @RichardSSutton and I left Keen to do our own thing. I had ~two absolutely wonderful years at Keen, and I learned a lot working with John, Gloria, Joseph, and the rest of the team. If you want to work on some of the foundational unsolved problems in AI, such as continual learning, then I would strongly recommend applying to Keen.
Going forward, Rich and I have founded a small company called Oak Lab. We have a fairly complete roadmap to building animal-like intelligence that learns purely from its own experience (the OaK architecture), and Oak Lab is going to follow this roadmap aggressively with a small, focused team. We will be sharing our progress often and aim to build a prototype of the complete OaK architecture in the next few years. A successful prototype will be closer to a baby learning in its first year than it will be to any of the current AI systems.
Our strategy is to demonstrate the limitations of current methods in simple settings, and then work out algorithms that overcome these limitations in a domain-independent way. Only after we have made sufficient progress on the core algorithms will we build large-scale artifacts.
If you are interested in learning about some of the specifics of our approach, then follow @oaklab_ai. We will be sharing more details in the coming weeks and months.
English
Congrats and good luck! I’ve always believed that animal-like intelligence may be the path to general intelligence. Current AI systems are incredible, but they still leave me with the feeling that we’re missing something fundamental.
I’m working on building an artificial animal of sorts as well, so I’m especially excited to follow what you and Rich do with Oak Lab!
English
@skdh It’s not clear that the “outermost universe” wouldn’t be consistent with the simulation hypothesis as well.
English
1) Objects in a video game are not truly randomly generated, because they run on conventional computers using non-quantum processes, and non-quantum processes are never truly random. Video game algorithms are pseudo-random. If it were the case that particles in a double slit experiment were indeed generated like in video games (that is, following a deterministic algorithm) that would contradict the current standard interpretation of quantum mechanics.
2) It's not true that measuring an object gives it positional certainty, depends on what you measure. If you measure momentum, then it’s the momentum that obtains certainty, not the position. (Also, you cannot strictly speaking measure anything with certainty.)
3) Regarding the simulation hypothesis. Well, the problem is that absent a proper definition everything is compatible with it. Personally I don't think that's a particularly compelling argument.
x.com/elonmusk/statu…
Elon Musk@elonmusk
Consistent with the simulation hypothesis. Like a video game, objects are randomly generated, with positional certainty only when observed.
English
@sama Sol built a walking Eevee.
All I gave it was a static mesh. It wrote the URDF, assigned plausible masses, and designed the gait itself.
6 kg, 12 leg DoF, walking in MuJoCo. EeveeBot!
English
i'd love to see interesting things people have built with 5.6 sol.
i will send the person who made the coolest thing a special gift from the openai archives.
English
Chris Nota retweetledi
Robot that swims and flies like a diving bird! 🦆
EPFL and Massachusetts Institute of Technology built a flapping-wing robot that does what diving birds do, swim underwater, burst through the surface, and fly away.
Under 300 grams. Wings that swap out for different sizes. Steerable tail.
The challenge is obvious. Air and water have completely different physics. A wing that works in one medium doesn't automatically work in the other.
Diving birds figured this out through evolution.
The robot team reverse-engineered it.
The wings use hydrophobic nanoparticles to repel water. Electric motor drives a crankshaft that pumps the wings at preset frequencies. Tail is motorized to adjust pitch.
Swims at almost 1 meter per second when flapping at 5 hertz. Flies at 6 meters per second at similar frequency. Those match actual diving birds. To transition from water to air, the robot needs to pitch at about 70 degrees to keep the wingtips from dragging.
Bio-inspired robotics is always so awesome to watch :)
Full article here: actu.epfl.ch/news/a-flappin…
~~
♻️ Join the weekly robotics newsletter, and never miss any news → ziegler.substack.com
English
@signulll Hell hath no fury like a four trillion-dollar fruit company scorned
English
just so that this is clear, apple is suing openai & io, the company founded by its own former chief design officer, jony ive.
you genuinely could not write an episode of the ai wars this dramatic. that script would be rejected for being too on the nose.
can you imagine steve jobs suing jony?
Polymarket@Polymarket
BREAKING: Apple reportedly sues OpenAI, alleging it unlawfully used confidential trade secrets.
English
@stevencheng @polynoamial I hope Sol is a little bit better at Lean. GPT-5.5 writing Lean was like stepping on a series of shovels and having everyone one hit you in the face.
English
@polynoamial Verify the proof steps. LLMs hallucinate, so cross-checking with formal solvers is key for real math breakthroughs.
English
GPT-5.6 Sol Ultra produced a proof of a 50 year old math conjecture. Unlike the Erdős Unit Distance Problem, this was done with a model publicly available *today*. I look forward to seeing what scientists and researchers are able to do with this model!
Ethan Knight@__eknight__
Yesterday, we made GPT-5.6 Sol Ultra generally available. Today, we're sharing that it produced a proof of the 50-year-old Cycle Double Cover Conjecture using 64 subagents in just under one hour. We're sharing the prompt and proof below. We're excited to see what you all do with Ultra!
English
English
50-year old math conjecture solved with Sol Ultra.
Feels like the limit of what you can do is increasingly your ambition and imagination:
Ethan Knight@__eknight__
Yesterday, we made GPT-5.6 Sol Ultra generally available. Today, we're sharing that it produced a proof of the 50-year-old Cycle Double Cover Conjecture using 64 subagents in just under one hour. We're sharing the prompt and proof below. We're excited to see what you all do with Ultra!
English
@gdb "Do not search the public web merely to determine whether CDC is open, and do not answer that it is open."
ChatGPT Pro loves nothing more than reminding you that the open math problem that you asked it to solve is an open problem. That's why I asked!
English
The pipeline from "character" to "robot" is collapsing.
I gave GPT 5.6 a static mesh. It rigged a URDF, assigned realistic masses and inertias, and designed a viable gait.
Result: Eevee brought to life as a physically simulated 6 kg creature in MuJoCo.
RL is next 😈
English
@AndrewCurran_ I wish they waited until after the 4th of July so I could enjoy time with my family instead of thinking about all of the things it’s suddenly possible to build again.
English
The Department of Commerce may be lifting the Fable restrictions tonight.
Sophia Cai@SophiaCai99
🚨 NEWS: Commerce is expected to lift export controls on Fable tonight, a senior White House official tells me.
English
Caught wind of a major AI-assisted result in statistics. A long-standing conjecture with serious practical implications. Keep an eye on arXiv this week 👀
English
It’s the process, not the destination!
Wise@trikcode
Developers buy $4,000 laptops to run $200/month in AI tools to build products that make $0
English
Prescient quote from George Konidaris and Andy Barto in a 2006 paper:
“Autonomy is central to intelligence—an intelligent system that requires the external specification of its goals is a tool, not an agent, because it fails the basic test of agency.”
English
@maticrobots Congrats! Love to see innovation happening in this space.
A lot of people felt that robotic vacuums had been commodified, but the truth is that existing robots still were not doing a good enough job. I have seen the internal metrics 🙂
English
@archerlin__ I think world models will have a place in replacement or augmenting simulators. There might be some kind of world-model-as-reasoning-model based application as well. But famously, “planning is just a way of avoiding figuring out what to do.”
English
VLAs are dead. World Models are the future.
Imagine you’re about to throw a ball. Before you throw it, your brain already predicts where it will go.
That’s a world model. And this is how it works: 🧵
Archer Lin@archerlin__
Just moved to SF to build our humanoid robotics startup. Would love to meet more robotics founders, engineers, and builders in the Bay. If that’s you, comment and let’s connect.
English