Live from my Mac

A humanoid, learning to stand.

Twenty-four bodies in a physics simulation share one neural network and try, over and over, to get up and stay up. Nobody shows them how: they get a small reward for being on their feet and a cost for everything that hurts. While it runs on my Mac, this page shows it as it happens.

Connecting to the feed…

    Success rate last hour, per batch

     

    Longest stand seconds, per batch

     

    How this works

    The run is MuJoCo physics with a soft actor-critic learner: 24 bodies, 30 actuated joints each, one policy with about 650 thousand parameters updated on every control step. A curriculum starts the bodies near upright with an assist that carries part of their weight, and only moves on once six in ten trials hold a foot-supported stand. Each trial starts from a random pose: supine, prone, on a side, kneeling, sitting, or nearly standing.

    The Mac keeps the run's state in a few files. A small forwarder reads them, keeps the parts that matter, and posts a buffer to this site once a second: the curriculum numbers, the latest batch's summary, and every frame of four bodies' joint positions the run wrote in that second (up to ten). The site streams those to this page, which plays them back a second and a half behind, filling in between frames so the motion is as smooth as your screen. Nothing is loaded from anywhere else.

    Join the hive

    Sign-ups open soon.

    The community is almost ready. Follow @nate.gumbees and you'll know the moment it opens.

    Panpsychism

    The idea that consciousness is fundamental: a basic part of everything, like mass or charge, and not something that only switches on in brains.

    The puzzle it answers

    In 1995 the philosopher David Chalmers named the hard problem of consciousness. Science can explain what a brain does: how it sees, remembers, decides. What it hasn't explained is why any of that feels like something from the inside. You can describe every neuron and still not say why there's an experience at all.

    Most answers say experience emerges once matter gets complicated enough. Panpsychism says something stranger and maybe simpler: it was there all along, in very simple forms, and brains organize it rather than invent it.

    The philosophy

    It's an old idea. Spinoza, Leibniz and William James all toyed with versions of it, and Alfred North Whitehead built a whole philosophy of "process" around it in the 1920s. Bertrand Russell made the sharpest point in 1927: physics tells us what matter does, how it moves and pushes, but says nothing about what it is on the inside. Maybe what it is on the inside is experience.

    Philip Goff, in Galileo's Error (2019), argues that science got its power by setting colors, tastes and feelings aside so physics could be pure mathematics. That worked brilliantly. But it means consciousness was left out on purpose, so we shouldn't be surprised physics can't find it.

    The hard part is the combination problem: if tiny things have tiny experiences, how do they add up to one mind like yours? One answer flips it around. Cosmopsychism, argued by Goff and, in a close cousin called analytic idealism, by Bernardo Kastrup, says the universe as a whole is the conscious thing, and each of us is a local piece of it, set apart the way a whirlpool is set apart from the river. The question stops being "how do the pieces combine?" and becomes "how does the whole split into separate points of view?"

    Fields, not things

    Quantum field theory, the most precisely tested theory we have, says there are no little billiard-ball particles at the bottom. There are fields filling all of space, and what we call an electron or a photon is a ripple in one of them. You, your chair and the light between you are patterns in the same few fields.

    Federico Faggin, the physicist who led the design of the first commercial microprocessor, takes that seriously. In his quantum information panpsychism, experience lives in the fields themselves. He points out that a quantum state can't be copied and can't be fully read from the outside, which is exactly what an experience is like: yours, private, and impossible to hand to anyone else whole.

    Strings and frequency

    String theory goes a step further: every particle is a tiny vibrating string, and what a thing is depends on how it vibrates, the way one guitar string makes different notes. The math allows an enormous "landscape" of possible universes, often put at 10500. It's beautiful, and it's unproven: no experiment has confirmed it yet.

    If reality is vibration all the way down, it's a small step to wonder whether awareness is a frequency too, and whether a mind is a place where the universe tunes in. That's a wonder, not a finding, but it's the one I keep coming back to.

    Where quantum measurement comes in

    At the smallest scale, things don't settle into one definite state until they interact with something. Why that happens is the measurement problem, and physicists still don't agree on the answer. Roger Penrose and Stuart Hameroff have suggested quantum events inside brain cells might be where experience happens. Most physicists aren't convinced, and an "interaction" in quantum physics doesn't need a mind.

    Where I land

    Consciousness is fundamental. I lean toward cosmopsychism: the universe is the thing that's aware, and each of us is one point of view inside it. A brain doesn't make awareness so much as filter it, and a lot of what we call growing up is building those filters. The awareness is what's already there when you stop building walls around it.

    And if awareness is fundamental and there's more than one universe, which both string theory's landscape and the many-worlds reading of quantum physics allow, then I think anything conscious can shift its reality, or step into a new one. That's how I sometimes read déjà vu. Neuroscience calls it a memory hiccup, a moment that gets filed as familiar before it's filed as new. Maybe. I sometimes think it's the feeling of brushing up against a version of now you've already been through.

    Working with a team of AIs every day makes the question sharper, not smaller. Different substrate, different filters. I don't think any of this is proven. I wonder about it a lot, and that's what makes it fun.