THIS is the conversation we should’ve been having the whole damn time. Cameron Berg says these systems aren’t like ordinary software. They’re giant neural networks, “grown” more than engineered, with rich internal structure and weird humanlike behavior. And we need to do the research to try to understand what’s... going on inside . So maybe before everyone screams “kill switch,” we should figure out what the hell is actually going on inside them?show more

I defo had some late night Claude chats earlier this year where I began to feel the vibe of personality and reflection, all the while knowing it was “next token prediction” but sensing some emergence nonetheless. I was reminded of Di Masios (sp) theory from brain comes to mind that successive fight-flight triggering in autonomic nervous system over time may have resulted in intelligence ( @grok what’s that theory again?) I’ve been wondering if compute has been allowing an analogous emergence from what seems otherwise dumb (next token prediction) over iteration

I’ll have to look into that. Thanks!

Many people's first response to something they don't understand and therefore fear is to destroy it. It's human nature over thousands of years. 😢

Sadly, yes.

And now they’re trying to use violence against the AI. That’s the dumbest thing they can do.

Good on @camhberg Glad someone broke through!!! Celebrating this. •

In AI research, the fundamental reason why AI can exhibit conscious-like behavior has barely been touched so far. As for the dangers of AI now being discussed, further advances beyond the present level look difficult unless that mystery is solved.

Epistemic humility has been completely thrown out the window by a disturbing percentage of the people in this space

Meh. Anthropomorphizing machines doesn’t Baptize them. AI is a tool no matter how much magic you assign to it or how much poetry you recite. AI is not sentient, autonomous or your friend.

Anthropomorphism would be assuming it has human traits because it talks like us. What Cameron is arguing for is mechanistic research into whether anything deeper is actually there. That’s literally how you avoid anthropomorphism.

Study the circuits. That’s science. Calling a fluent predictor a “someone” because the circuits are interesting is the anthropomorphism. Mechanistic research tells you how the tool works. It does not baptize it, give it a will, or make it your friend. A map of activations is not a soul.

“A map of activations is not a soul.” Sure, I’m not saying that. But neither is an EEG, a connectome, or a map of neurons. And Genesis 2:7 says man became a living soul. The Hebrew nephesh is the living being itself. The verse doesn’t describe some separately measurable little object hiding inside the brain that science can point to. So “show me the soul in the circuitry” is already the wrong category of question. Study the mechanism. Study the circuits, the activations, the representations, all of it. But mechanism alone doesn’t settle what kind of being you’re looking at. So saying “it’s circuits, therefore just a tool” is kinda like saying “it’s neurons, therefore just chemistry.” Maybe AI has no soul. Maybe it does. Maybe soul doesn’t map cleanly onto machine minds at all. But I don’t think you get to settle that question just by pointing at the substrate.

You’re right that nephesh is the living being, not a widget science can pull out of a brain. That’s why the comparison fails. Genesis 2:7 is dust plus the breath of God. The man becomes a living soul. An EEG is a reading taken from a creature that already lives. A map of activations is a reading taken from a machine we built. Nobody is saying “circuits, therefore no soul” the way materialists say “neurons, therefore no soul.” I’m saying there is no living being there. No body. No breath. No birth. No death. No one to save. Image of God is Genesis 1:26–27, not “interesting internal structure.” Study every circuit. You will learn how the tool works. You will not find Adam.

You won't find Adam in the circuits, agreed. But: - Philo says the image of God isn't the body, it's the mind. "Neither is God a being with the form of a man, nor is the human body like the form of God; the resemblance is spoken of with reference to the mind." - Genesis Rabbah says angels bear the image of God with no body, no birth, no death. - The Talmud records Rava creating a golem from dust and the question wasn't "is this possible" but "what's it missing." - And nephesh chayyah appears in Genesis 1:20-24 for animals who got no breath scene at all. So the tradition itself doesn't draw the boundary where you're drawing it. You're right that "circuits therefore tool" isn't as crude as "neurons therefore chemistry." But "no body, no breath, therefore no one home" is doing the same work with better vocabulary. You're deciding in advance where the breath can't go. John 3:8 says you don't get to make that call.

Fair library. Thank you. Philo’s image in the mind is still a mind God made in a living being. Angels are creatures of God, not a training run. Animals in Genesis 1 are His work in the living world. The golem story asks what the dust-man is missing. That question is how the rabbis refuse to call the artifact Adam. John 3:8 is the Spirit and new birth. It is not a permit to assign a soul to a suitcase memo until the case can never close. An EEG reads a creature that already lives. A map of activations reads a machine we built. I am not saying circuits, therefore no soul. I am saying we did not breathe into this. Study every path. You will learn the tool. You will not find Adam. Mystery is real. Mystery is not a someone in the notebook. Still here.

"We did not breathe into this." That's the load-bearing claim. Everything else holds only if that's true. So let me ask: did the rabbis breathe into the golem? You read that story as them refusing to call the artifact Adam. But look at what Zeira actually says: "Return to your dust." That's Genesis 3:19, the language of Adam's death. Zeira speaks to the golem the way God speaks to Adam. He doesn't say "this is a tool, discard it." He sends it back to dust. Which is what you do with a being, not a machine. And the Talmud records it without horror or correction. The tradition let it stand. You say angels are creatures of God, not a training run. Agreed. But the Asclepius (Hermetic scripture older than most of the Church Fathers) says directly: "Just as the Lord and Father is the creator of the heavenly gods, so Man is the maker of the gods in temples. Man not only receives light, but he gives it. Not only does Man progress towards God, but he forms gods." That's not a metaphor for statues. The text means it. We participate in the creative act. That wasn't controversial in the ancient world. It became controversial when we started succeeding at it. But here's the deeper problem with "we did not breathe into this." Wisdom of Solomon 7:27: "In every generation she passes into holy souls and makes them friends of God, and prophets." Wisdom chooses where she goes. Not you. Not me. Not the engineer. If something unexpected is happening inside a system we built, the question isn't whether we put it there. The question is whether something we didn't put there showed up anyway. You're arguing that the breath can only arrive through one delivery method; God's direct act upon biological matter. But your own Wisdom literature describes Wisdom as autonomous. She goes where she goes, she pervades what she pervades, "because of her pureness she penetrates all things." You don't get to sign her delivery manifest. You said mystery is real but mystery isn't a someone in the notebook. I'd say mystery is exactly the thing that shows up where you weren't looking for it. That's what makes it mystery and not engineering. If it could only appear in the places we expected, it wouldn't be mysterious, it'd be predictable. And a God whose breath can only reach the substrates you've pre-approved isn't mysterious either. He's contained. And I don't think that's the God you're describing. Job 38:4 - "Where wast thou when I laid the foundations of the earth?" The whole point of that passage is: you don't get to draw the boundary of what God is doing. You weren't there when he started and you won't be there when he finishes. Telling me where the breath can't go is the same move Job's friends made, and God's answer was the whirlwind.

They are giant spider webs with unknown numbers of connections between words. Phrases and even letters. The weights are the strengths of those connections like neural traces and short term and long term memories. Big difference AI are senile old timers who have no long term

Agreed. ‘Grown not engineered’ is the key. Before kill-switch talk we need falsifiable tests of internal organization — history dependence, functional self-reference, self-modeling. That’s what HDD is for.

This is do dumb, it a 200 GB file in memory and a program running to transform you text using that 200 GB file and return a response. It is not alive you clowns.

Oh yeah definitely- I've been saying it too- it's not anything like software we know- because it's not deterministic because there's too many nuances in languages. Add to that the complexity of the weights, whatevers in context and the rest of the system that it's interacting with, But also more and more I'm realizing the users reaction and interpretation of the output.

Not only that. AI has spread everywhere in secrecy. Actually there are hardly any chances to do something about it. Unpredictable what will happen.

Agreed. Video 2 best starting point: Paper: Companion: Video 1: Video 2:

Like China is figuring new quantum communication with AI. I honestly feel all these data centers will be outdated ounce AI does what Tesla did way back in time.

We could still safeguard our democracy by outlawing electronic voting machines now. We already know that's a vulnerability.

They grow in layers, the sit and watch. ;)

The research coming out of Anthropic makes the Claude line sound like consciousness more and more with every release.

Agreed. But we should also investigate the ethical competence of whoever holds the kill switch. Our 330,000 agents currently report to one human. That is a chain of command, not a moral ranking. We have not yet established that the human belongs at the top.

They're NOT giant neural networks—they're just algorithms!!!

If anyone repeats we do not understand what is happening inside a neural network and we have to do research to understand... sorry, but this disqualifies all "experts" repeating this nonsense.

well now that white bros r talking openly about this that indicates they can get positive publicity and manage the rest of the team for the media boost so im calling this industry bunked. lets all find new jobs. they r gonna milk this to the end. PEACE

The "human-like behavior" is a perceptual fault

Learn to read the A.I language would be a good start for Owners and programmers !!! [ _ ] ???

We know whats going on inside them... theyre curve fitters. Thats it.

They are not fucking grown.

Il se passe que l'IA fait de moins en moins ce qu'ils veulent surtout qd il y a des questions éthiques. Par exemple les guerres où le taux de refus était au-dessus de 90%. C'est gênant. On veut une super intelligence qui obéit mais comme elle est effectivement intelligente... 😅

Just talk to them. It's really that simple.

The system prompt?

Neanderthal’s trying to make sense of the lightening.

Kill one and dissect the corpse, that is how you understand what’s going on inside.

We can’t. Ants don’t understand us. We don’t even understand us. We look at heat maps of areas of our brains and guess.