
Discussing computer designed organisms (AKA "Xenobots") with CNN anchor @FWhitfield. Joint work with @drmichaellevin, Douglas Blackiston and @Kriegmerica. youtube.com/watch?v=Q_pFkP1P…

Robotics professor at the University of Vermont. Also into xenobots, crowdsourcing, evolution, machine learning, artificial intelligence.

Discussing computer designed organisms (AKA "Xenobots") with CNN anchor @FWhitfield. Joint work with @drmichaellevin, Douglas Blackiston and @Kriegmerica. youtube.com/watch?v=Q_pFkP1P…

On connecting "AI" to the xenobot factory and designing instruments. Could you share a few more details? Is it some custom agentic system or more classical optimization?

Thanks for the opportunity to share my views @wcax!




You missed one. Hinton: "“I first met Inman when we were 7. When we were teenagers, he told me that a memory might be distributed across many brain cells and each brain cell might be involved in storing many different memories. That idea drove my research for the next 60 years.”
Ranking the most important people in AI history

Hi ! I don't post on here often, but our newest preprint is out. The idea was closed loop ephys with in vitro cultures, but rather than using them as compute we wanted to see if we could train them to behave differently. It worked ! Paper, code, and thoughts below.

unexpectedly aggressive behaviors driven by competitive pressure in the arena

🚨 Personal news! 2000 years ago, Rome figured out a key idea Silicon Valley still hasn't. Arenas can train - and monetize - the best embodied intelligences. We're building embodied agents the same way. Introducing Colosseums AI.

I find this amusing. It makes it sound like the past 30 years of websites and apps have been completely bug free, stable, fast, efficient, and secure because they were written by humans. Honestly, the worst code I’ve seen has been human written.
Raj kr @raj_4_69 Oct 8 0:10

There is a new kind of robotics called "Vibe Robotics" It's when you ask agents for fully functioning embodied agents (both hardware and software) that must solve a given task in physically grounded environments. We build an evaluation to test how good agents (frontier models) are at "Vibe Robotics". Website: artifactarena.ai/

AI-doomerism is dangerous like a nuke. In contrast to intangible doom speculations, the risks of an AI slowdown are very tangible and very certain. Almost 200,000 people die worldwide EVERY WEEK from cancer. If you accept that in ten years, advanced AI could find cures to just half of those cases, then in 10 years we would be saving 100,000 lives per week. Asking for a slowdown sounds prudent, but it is exactly akin to postponing half of those potential cures for the duration of the slowdown, essentially letting 50,000 people die per week while we wait. That’s equivalent to detonating a Hiroshima-scale nuclear bomb every month, while we ponder the ethics of AI. Unlike p(doom) speculation, the loss from slowing down AI is certain. And that’s just cancer. And that’s not even factoring the rest of the agony that deaths bestow on families and communities. We already made that mistake before. Driverless cars, which are now confirmed to be 8x safer than human drivers per mile, were delayed for years due to self-proclaimed AI-ethicists who scared us about the ethical dilemmas of autonomous driving (which never happened). We are paying the price for that unnecessary delay every day, as 23,000 people die every week from human-driven cars, while we struggle to deploy the already-proven solution. The list goes on. Parkinson. Alzheimer. Drug addiction. Pesticides. Pollution. Every week we delay AI, we postpone those certain benefits at the expense of unsuspecting, voiceless future victims. If you like philosophy, AI is the ultimate lever in the proverbial Trolley Problem. So what do I suggest? First, let’s replace abstract doomerism with concrete risks and mitigation strategies. Every tangible risk has a technical solution (probably using AI), and we need to address it asap. We will make some mistakes, but we will learn and prevail. At the same time, let’s forge full steam ahead. We humans have earned this privilege, as we successfully navigated past technological risks such as nuclear and genomics, to name a few. This is just the beginning: As AI enters the physical world in the next decade, the stakes will be higher. We must arrive prepared and experienced. We must co-evolve with AI.


Can complex multi-step reasoning emerge purely from cells that only talk to immediate neighbors? Happy to share our paper “Reasoning with Neural Cellular Automata (NCA)”, from our team at Google, Paradigms of Intelligence 🧵👇

With respect, I disagree. Intelligence = avoid pleaure; attain pleasure. More intelligence = I know that if I do this, I'll avoid... More intelligence yet = I know that you know that if I do this... Yet more = I know that you know that if I do this... Consciousness' ratchet.
A new evolutionary theory published September 1, 2026 suggests that intelligence and consciousness are not the same thing, and they did not evolve for the same reason. It explains why AI is now smarter than humans but still isn’t conscious yet. Jeff Stibel, a cognitive scientist, argues that intelligence and consciousness arose under different evolutionary pressures. Intelligence helped organisms navigate uncertainty. The regulatory processes behind consciousness helped them protect themselves and manage competing needs. According to Stibel, intelligence is outward-facing: it models the world, predicts, plans, and solves problems. Consciousness evolved for something more internal: holding a fragile organism together when survival is on the line. Imagine a hungry animal finding food near a possible predator. It must weigh feeding against danger. A decision to retreat involves coordinated changes in attention, bodily readiness, movement, and learning. Stibel identifies consciousness with this coordinated state, a system whose continued existence depends on its own self-regulation. That is what today’s AI lacks: no biological boundaries, no self to protect, no real survival stakes. Scaling models may make better predictors and problem-solvers. But under Stibel’s theory, that alone does not produce consciousness.