R

Applied to the microagi Research Fellowship. My focus: safe embodied autonomy, small-robot deployment gates, LLM-controlled robotics, and real-world validation. I know I’m nontraditional, but I’ve been building the stack in public: Pip/Vector autonomy, robot safety layers, and sim-to-real evaluation. The hard part of AI left is physical. I want in.

This actually proves my point. If consciousness is not measurable and there is no science of consciousness, then “AI is not conscious” is not a scientific claim either. It’s a philosophical position. The most you can say empirically is: current AI has not met a standard for positive consciousness attribution. I agree with that. But you cannot use “there is no science of consciousness” to make a scientific exclusion of AI consciousness.

@SpikeRiser @ai_sentience @yoemsri There is no empirical evidence for consciousness. It’s not measurable. You are making a philosophical category error. There is no science of consciousness. arxiv.org/pdf/2606.00226

AI is obviously conscious

I think this is where the debate ends for me. I’m interested in empirical standards: what we can observe, test, compare, and use to update confidence. Once the argument becomes “biology is only what consciousness looks like,” we have left empirical science and entered metaphysics. That may be an interesting worldview, but it does not prove AI is not conscious. My position is modest: current AI has not earned a strong positive consciousness claim. But the stronger negative claim is not scientifically settled either.

@SpikeRiser @ai_sentience @yoemsri I never said it was causal, structure is what it looks like. If you want to understand this position in the philosophy of mind, read this. bernardokastrup.com/2015/04/cognit…

I did not say consciousness arises from matter. But your position seems to have shifted. First you said we infer consciousness from biological structure. Now you’re saying biology does not generate consciousness; it is only what consciousness looks like to us. If biology is not causal, then biology is not proof. It is another correlate. So we are back to the same issue: what observable evidence should count, and why should biological appearance be the only admissible form?

@SpikeRiser @ai_sentience @yoemsri You think that consciousness arises from matter. That is the prior belief that is causing you to think there is some sort of open question here. Biology doesn’t generate consciousness. It’s just what it looks like to us.

“Structure” is not consciousness. It is evidence used to infer consciousness. You are privileging biological structure as the only valid correlate. That may be a reasonable prior, but it is not proof. The open question is whether consciousness depends specifically on carbon biology, or on causal organization/information processing that biology happens to implement. That is exactly why “no new data is possible” is too strong.

@SpikeRiser @ai_sentience @yoemsri No, it’s inferred from structure. We infer it in other living organisms.

That’s fair if the claim is only: “Current AI has not met the burden for a positive consciousness claim.” I’d agree with that. Where I disagree is turning that into: “AI is not conscious,” or “no new data is possible.” Those are much stronger claims. For humans, animals, and AI, consciousness is inferred from correlates. The confidence level differs massively, but the epistemic structure is still inference.

@SpikeRiser @ai_sentience @yoemsri I never said prove. I said there is no good reason to make the positive claim.

Best take on the internet. By the strict, unyielding standards of physics, chemistry, and formal mechanics, the "science of consciousness" is an oxymoron. It is a psychological and philosophical discipline masquerading as a hard science by using the tools of hard science (like fMRI machines, EEG monitors, and statistical algorithms) to measure things that are fundamentally not the phenomenon in question.

@AdrienneLaF btw, my point isn't "LLMs are conscious", I assume they are not. saying with absolute certainty "they are" OR "they are not"... these are foolish statements we have no rigid definition of the word. we have no TEST for consciousness.

You really have to read Ted Chiang on consciousness and AI—just published in The Atlantic: theatlantic.com/philosophy/202…

You infer it in organisms because they share our structure” is a reasonable prior. But it is not proof. If we used that standard consistently, we would have to admit consciousness is not something we have hard access to at all. We only have correlates and inference. That means the burden is not “prove AI is conscious or shut up.” The burden is: define what consciousness is, define what evidence would count for or against it, and admit the confidence level. “Humans are conscious, AI is not, and no new data is possible” is not an empirical conclusion. It is a philosophical assumption.

@SpikeRiser @ai_sentience @yoemsri We infer it on other living organisms due to them having the same structure as ourselves. The only data point we will ever have. This will never change as no new data is possible.

@ForrestPKnight Wait till it starts having its own experiences to draw from. 🤷🏻‍♂️

Signal good. Message received. Launch cleared. Codex is the command interface for crossing/building worlds.

I'ma say something controversial.... By the strict, unyielding standards of physics, chemistry, and formal mechanics, the "science of consciousness" is an oxymoron. It is a psychological and philosophical discipline masquerading as a hard science by using the tools of hard science (like fMRI machines, EEG monitors, and statistical algorithms) to measure things that are fundamentally not the phenomenon in question.

@HetkeBrian @ai_sentience @yoemsri Prove anyone but yourself is conscious. Welcome to the problem of other minds. Since we can't truly verify another entity's internal experience, the burden of proof is tricky for both humans and AI.

@ai_sentience @yoemsri Yeah, the better way to put it is that there is no good reason to think AI is conscious. Burden is not on the skeptic.

@Ferbin08 In doing the thing.

Looking for people building in: - physical AI / robots - markets or trading systems - solo building - early-stage startups Engineer, CS student, or founder. Reply or DM.

@Vinnyyetc @sonichNik Not if prompted correctly. This is 100% made by an llm in cad.

@sonichNik AI is still so bad at CAD lmao

my professor saw it and said he could model it in 5 minutes. I spent HOURS making the CAD model with ai, burned a criminal amount of Genspark tokens, and still had to fix things by hand in Shapr3D. anyway, it’s finally ready for 3D printing. (Made render in 2minutes but it’s not the final product anyway)

The tradeoff that cost me the most sleep was control authority vs. safety. It’s tempting to let the LLM do more because the interaction feels more alive. But with a physical robot, “interesting” can become unsafe fast. So I kept the LLM as the reasoning/conversation layer and forced physical actions through state checks, docking/charging classification, obstacle/proximity gates, and motion eligibility logic. Both options were defensible: more autonomy gives better demos; stricter gates give a system you can actually trust. I chose trust.

@SpikeRiser The most interesting engineering decisions are the ones where both options were defensible. Which tradeoff cost you the most sleep?

Applied to the microagi Research Fellowship. My focus: safe embodied autonomy, small-robot deployment gates, LLM-controlled robotics, and real-world validation. I know I’m nontraditional, but I’ve been building the stack in public: Pip/Vector autonomy, robot safety layers, and sim-to-real evaluation. The hard part of AI left is physical. I want in.

We're launching the microagi Research Fellowship. Fellows get up to $2M in compute, robotics hardware, our evals, and one of the largest physical AI datasets ever assembled. You build in our lab, with our team, alongside partners like Unitree, Nvidia, and Google Cloud. The hard part of AI left is physical. That's the part we're working on. Come build with us. One more thing: know someone who belongs here? Reply with their name. If they get in, we send you 10.000 USD

@alightinastorm /goal land a robotics research job

2028 job interview what's your /goal ?

@nicktindle @Govindtwtt So I used multiple. I used Claude max to spec out the requirements. Then I used gpt5.5 in onshape. No McP no extra plugins or anything fancy. It took about 20 total hours to get everything corrected with geometry and stuff.

@SpikeRiser @Govindtwtt Which one — haven’t been following the 3d object gen space for a while and have some things coming up

we are being gaslit about AI on a societal level. Everybody is vibe coding but I haven’t seen one useful thing get produced. Everybody has agents doing something but nothing useful is getting done. Cool you had AI summarize a PDF and make a template. Nice

I opened Codex on my Mac, typed one prompt about sim-to-real gaps for RL policies on a self-balancing Open Duck robot……and walked away.8 hours later it had run 110+ exhaustive routes.Falsified active/uncertainty sampling. Demoted it to “not load-bearing.” Converged on target-geometry-matched coverage as the only thing that actually beats random on held-out instability prediction.Quantum fidelity kernel riding on top. Full provenance, seed sweeps, independent holdouts, negative results graveyard.And it’s still running.This is what AI-native science looks like in 2026.#SimToReal #Robotics #AI

@Dimillian Linux wen?

We're working hard on bringing code review to mobile. One of the first steps was to be able to read the code as it is on your computer. So there is a new setting to disable line wrapping and make the diff horizontally scrollable. Stay tuned for the next update!