Dive into the future of computing with Malcolm and Simone Collins on Based Camp! In this mind-bending episode, we explore the breakthrough in wetware—using real human neurons grown from skin and blood cells to power affordable bio-computers. From Cortical Labs' $35,000 neuron chips that play Doom to mini-brains mimicking kindergartners' neural patterns, we discuss how this tech is cheaper and more efficient than traditional silicon systems. We tackle ethics (including pain pathways in lab-grown brains), AI alignment, quantum integration, cultural perspectives from Puritan roots, and wild speculations on space-faring brain ships, human uploads, and a networked species beyond humanity. Is this the end of worst-case AI scenarios or the dawn of servitors? Plus, thoughts on techno-puritanism, Soma-inspired horrors, and why backwoods traditions embrace utility over mysticism.

The X posts we mention in this podcast:

Episode Transcript

Malcolm Collins: Hello Simone. I’m excited to be here with you today. Today we are going to be discussing a breakthrough that I hadn’t expected which is that. Using neurons in bio-inspired systems is now a reality that you, a watcher of this show can likely afford yourself. If you wanted to try some sort of like business experiment based on this, what and in many ways is now cheaper than doing it on computer.

And this was a huge breakthrough that changes a lot of, if you’re looking deep future of where humanity goes at this point. Mm-hmm. With the development of quantum computers, was the development of AI continuing one thing that a lot of people feared and, and this is why I say that. This is such a, like, a lot of people are like, Malcolm, this is horrifying.

Like, are you excited about servs and everything like that? Like humans being turned into like. Husks for a [00:01:00] machine,

Speaker 2: Define the damage. Spine. Compromised. Have you not received pain? Suppressants suppressing pain? Damage submitted report to the surgical bay.

Malcolm Collins: And it’s like, well, we’ll we’ll get to that, we’ll get to that. But what makes it really good is it changes worst case scenarios. Worst case scenarios for ai, foaming taking over the world, expanding into space.

Historically speaking before today I would say that in such a scenario as that, you know, humanity gets wiped out there is maybe a 3% chance that neurons or biological matter is part of whatever AI’s become. We are now, like if we’re using AI estimates here, because I was going through ai, having it compile all the research we have on where quantum computers are right now, you know, looking at computers a hundred years from now without humans around anymore it said 60 to 70% chance [00:02:00] that it would be partner on.

Simone Collins: Wow.

Malcolm Collins: So that’s, that’s now the worst case AI scenario, right? Mm-hmm. Likelihood this is, you know, humanity wiped out or enslaved our overlords. And, and what’s interesting is that the part of, and we’re gonna go into, okay, 50, 60 years from now, we project technology moving forwards and sort of the jumps that we’ve been seeing, technology moving forwards, what does a computer look like?

You know, quantum computing is working. We continue to see advancements in silicon-based computing. And we see these startups and companies continue to develop at this rate. Was it, was it neural computing? Yeah. What we’re gonna go into is, is, is what that computer is going to look like. Um hmm.

Speaker 15: , that does not mean the value of your existence turns negative to the contrary.

When it comes to the macro management of the civil system,. Your role has simply changed. Only. This can solidify the health and prosperity of future human [00:03:00] society,

Malcolm Collins: and what is, what is I think going to surprise a lot of people about what that computer will look like is it’s not gonna look that different from the ways that humans interact with computers today. By that, what I mean is the types of stuff that the quantum computer part of a brain made up of silicon neurons and quantum computers are going to handle is going to be very similar to the type of stuff that it would handle today.

Large scale logistical planning sort of stuff. No human is actually doing that with neurons. It’s just not the type of problem that we’re good at doing. Mm-hmm. The type of stuff that the neurons are gonna be doing is well, we’ll get to it, but it’s the type of stuff that actually humans do today within this arrangement.

The type of stuff that the silicon component is gonna be doing is the type of stuff that LLMs do today in this arrangement.

Simone Collins: Oh. It’s a perfect match.

Malcolm Collins: So we’re already sort of there already. Yeah. Yes. It’s, it’s very interesting. The, [00:04:00] the stuff that quantum computers are really good at mm-hmm. Is almost sort of opposite the stuff that neural arrays are really good at.

And so, yeah, let’s go, let’s go into the tweet that you sent me that prompted this. And we’re also gonna go into you know, the ethics of all of this. Why it’s ethically so cool. So awesome. Don’t, don’t be so squeamish about this guys.

Speaker: From the moment I understood the weakness of my flesh, it disgusted me icra for strength and certainty of steel. I aspired to the purity of the blessing machine

kind, claim flesh, as it’ll not decay and.

One day, the biomass [00:05:00] that.

Simone Collins: And had tip to not Alvis Huxley for sending this to us.

You rock.

Malcolm Collins: Yeah. Okay, so the tweet goes let me explain what just happened because I don’t think people realize how insane this is. Cortical Labs just put 200,000 real human brain cells in a silicon chip and train them to play doom in just one week. Each CL one system costs $35,000. So that’s affordable for, I mean, it’s expensive, but it’s not like a quantum computer or something like that.

Like if you had some business idea and you went to the bank, you could raise enough money to buy a few of these and operate them. Right?

Malcolm Collins: And one of the things I really wanna get into is the [00:06:00] cost, cost efficiency of these systems at their, at their most nascent stage versus existing systems that we operate LLM on.

And, and where they can do better and where they can do worse. And where we’re already seeing integrated systems that are doing things a thousand times cheaper than nonintegrated systems, which is really cool that we’re already seeing this. So a rack of 30 units consumes 850 to a thousand watts combined.

The human brain operates on 20 watts. So, so I wanna point out what this means here, right? For all of the calculations I’m gonna give you that are like right now you know, the, the neural systems are operating at, you know, one, 1000 subfraction of the silicon-based systems, right? If, if we’re, if we’re talking about their efficiency, because that’s what an AI that’s taking over the world or whatever is gonna care about this is what far future humans, when we’re building our giant brain ships, are gonna care about.

Because, you know, our, our, the, the, the, when you’re talking about like [00:07:00] space fairing systems you’re almost always gonna have like one super brain within a ship that I, I assume that this is probably the way that things are gonna work which is gonna be a network of some of the most advanced intelligences that you would have.

And then you will have, you know, microchips on phones and stuff like that. If people can say why I would say this. So if you look today one of the reasons that you have you, you don’t see this as much is because there is an intrinsic decentralization in the way that we use computers today due to distances, personal ownership, everything like that.

But if you have a, a space fairing ship the, there’s, there’s going to be, economic reasons to one, want the best brain on the ship to be one that’s powering your navigation systems. One that’s powering the decisions when the captain is asking an AI something, one that’s powering that one that’s powering the projections for the colony and everything like that.

But in addition to that, because you don’t have this huge amount of distance and everyone to an extent is going [00:08:00] to be working on behalf of the ship or of the early colonies it just makes sense to me when I’m asking my personal LLM on my phone, why not just outsource that to the ship based system?

So we’re gonna see a lot more centralization when we have space colonies and space travel than we see within existing systems. Mm-hmm. Which is why it makes sense to think about what do, what do these far future systems look like? But anyway, the point I’m making here when you’re thinking like, okay, where, where do we have neural tissue operating this stuff 30 of these.

Racks, which are a you know, a a a sort of like a, a single small chip, right? Single silicon chip. They take 850 to a thousand watts to run. Whereas the human brain operates on 20 watts. And what this means, well, that’s a

Simone Collins: difference.

Malcolm Collins: Yeah. There’s a huge efficiency gains to be gained here, right? Can we get more efficient than even the human brain?

I, you know, I think probably but at least what it means within the early days, if we’re looking at the other analog we have, the human brain is significantly more complicated than one of these [00:09:00] chips or a rack of 30 of these chips. So lots of, lots of advancements we can make to this. And. When we’r...