Wetware Computing Dr. Fred Jordan On Brain Cells, Biocomputers & The Future of AI

Summary

---

Podcast Summary: The Good Question Podcast

Episode: Wetware Computing — Dr. Fred Jordan On Brain Cells, Biocomputers & The Future of AI

Host: Richard Jacobs
Guest: Dr. Fred Jordan (CEO and co-founder of AlpVision & Final Spark)
Date: June 18, 2026

---

Episode Overview

This episode delves into the cutting-edge field of "wetware" computing—the use of human-derived brain organoids (mini-brains) as computational substrates for AI and potential future biocomputers. Dr. Fred Jordan, a leading entrepreneur in this space, challenges the current silicon paradigm and explores how biological systems could revolutionize efficiency, power consumption, and the architecture of computation.

---

Key Discussion Points

1. What are Brain Organoids?

• Definition and Clarification ([00:55]–[01:26])
  • “Mini-brain is not really realistic because it’s very far from a brain.” — Fred Jordan [01:16]
  • The correct term is brain organoids: collections of brain cells cultured to form simple, sphere-like structures, with varying degrees of structural and cellular complexity.
  • Organoids can be composed of a single cell type or a mixture; some have guided differentiation for more advanced functionalities.

---

2. Why ‘Wetware’ Computing? Biological vs. Silicon Brains

• Energy Efficiency & Biological Superiority ([02:16]–[04:10])

  • Human brains operate astoundingly efficiently: ~20 watts for billions of neurons and trillions of synapses. In contrast, simulating this biologically-inspired architecture on silicon demands the power of a small nuclear plant.
  • “Your brain runs on 20 watts, but you have 100 billion neurons, 10,000 synapses per neuron… if you wanted to make a simulation… will take like a small nuclear plant.” — Fred Jordan [02:27]
  • Biological brains are fundamentally distributed systems, with each neuron as a computer, unlike centralized, bottlenecked GPU/CPU infrastructure.

• Integration with AI and AGI ([04:11]–[05:07])

  • The “hybrid” approach: Combine language-based LLMs (silicon) with wetware systems for richer, more generalized intelligence.
  • “Maybe LLM is not going to be sufficient to reach AGI… One approach would have a hybrid approach… both being integrated together… Digital world and organic world like wetware.” — Fred Jordan [04:10]

---

3. AI, Reasoning, and the Brain

• The Limits of LLMs ([05:08]–[06:30])
  • Host observation: LLMs appear to merely “interpolate in latent spaces”—they don’t truly “think.”
  • Dr. Jordan: “Are we way better than this? So for sure we are not manipulating words when we are reasoning. Maybe actually we don’t need words… It’s not even sure that the brain was made for thinking… The brain is actually made for reacting and processing information, not thinking.” [05:39]
  • Biological and artificial systems represent fundamentally different approaches to processing information.

---

4. Experimental Progress: Building and Using ‘Wetware’ Computers

• Power Efficiency: Orders of Magnitude Better ([06:30]–[07:21])

  • Biological computation may be up to a million times more energy efficient than silicon, though practical implementation may only achieve 100–1,000x improvements due to technical constraints.

• Remote Experiments and Accessibility ([07:22]–[07:52])

  • “We do it on a daily basis… put them on electrodes and make them available for researchers… access it remotely in Python…” — Fred Jordan [07:31]
  • Researchers can program and interact with organoids remotely, stimulating them, reading outputs, and modulating biochemically.

• Connecting Multiple Organoids: Assembloids ([08:00]–[08:45])

  • “We made organoids of striatum, hippocampus, and cortical neurons … put them with daisy chain… they will fuse together in a matter of days.” — Fred Jordan [08:04]
  • These networks potentially mirror aspects of biological memory, computation, and reward.

---

5. What Are Wetware Computers Used For?

• Not for Traditional Computing Tasks ([08:59]–[09:13])

  • “You don't want to use wetware computers for number crunching… you’re not going to be happy.” — Fred Jordan [08:59]
  • The key application is AI research—particularly where biological realism is advantageous.

• Potential Scale: Organoid-Powered LLMs? ([09:18]–[10:37])

  • Order-of-magnitude estimates: one biological neuron can effectively replace thousands of artificial neurons in some contexts.
  • “It's absolutely possible to recreate a ChatGPT [with organoids]… way less neurons… Yes, I think it’s absolutely possible.” — Fred Jordan [10:20-10:37]

• Limitations and Scaling Challenges ([11:18]–[13:00])

  • Each organoid is ~10,000 neurons; to scale for production servers, larger neuronal masses (“neuronoids”) are envisioned.
  • The biggest limiting factor in scaling up: vascularization—keeping larger tissues alive with proper “blood” supply.

---

6. Practical and Ethical Considerations

• On Using Animal Tissue vs. Stem Cells ([11:38]–[12:33])

  • Dr. Jordan’s group chooses induced stem cells to avoid ethical issues and achieve scalability: “Much more convenient, I can tell you.” — Fred Jordan [12:34]
  • This allows for mass production of neurons without sacrificing animals.

• Dreams and Future Milestones ([15:02]–[15:26])

  • Long-term goal: scale up to “become a serious competitor to Amazon Web Services.”
  • Immediate focus: mastering the training and scalability of these alvolums (organoid volumes).
  • “We have to master training… scale up… to become a serious competitor to Amazon Web Services. We are going to do it.” — Fred Jordan [15:15]

---

Notable Quotes & Memorable Moments

• On the astonishing efficiency of brains:
  “If you wanted to make a simulation of this, that means basically run on a computer, traditional computer, your own brain will take like a small nuclear plant, you know.” — Fred Jordan [02:27]
• On the purpose of brains:
  “It’s not even sure that the brain was made for thinking. Okay. The brain is actually made for reacting and processing information, not thinking.” — Fred Jordan [05:39]
• On the vision for wetware computing:
  “It's absolutely possible to recreate a chatgpt [with organoids]… Yes, I think it’s absolutely possible…” — Fred Jordan [10:20–10:37]
• On moving from organoids to larger tissue masses:
  “Big is easy, but big and living is more complicated due to the blood supply.” — Fred Jordan [12:48]
• On the company’s ambitions:
  “We have to master training of the alvolums. We have to scale up in order to become a serious competitor to Amazon Web Services.” — Fred Jordan [15:15]

---

Timestamps for Important Segments

• Definition & distinction of organoids / mini-brains: [00:55]–[01:33]
• Why wetware is efficient / biological vs. silicon: [02:16]–[04:10]
• Host and guest on limits of LLMs and biological reasoning: [05:08]–[06:30]
• Practical experience developing and deploying wetware: [07:22]–[08:45]
• Potential for large scale / scaling limitations: [11:18]–[13:00]
• Company’s focus and vision for wetware biocomputing: [15:02]–[15:26]

---

Further Information

• Dr. Jordan’s teams at AlpVision and Final Spark are pioneering the commercialization and scientific deployment of wetware computing platforms.
• To follow their work, join their Discord via their company website.

---

This summary captures the essential discussion points, insights, and unique perspectives from Dr. Fred Jordan’s interview on the Good Question Podcast, offering a foundational understanding of the exciting future of biocomputation for listeners and readers alike.

Transcript

[A]

You're listening to the Good Question podcast with Richard Jacobs. Our goal is to make each of our guests exclaim, hmm, that's a good question. I don't know the answer. Because when that happens, it means you, the listener, may be inspired to learn more beyond the interview and to ask great questions yourself that lead to new insights. In this podcast, we cover historical and current anthropology, comparative religion and history. Welcome. And let's get started.

[B]

Hello, this is Richard Jacobs with the Food Question podcast. My guest is Fred Jordan. He's the CEO and co founder of AlpVision and Final Spark. We're going to talk about what's called wetware. Scientists are using human mini brains to actually power computers. So it should be a very interesting topic. Welcome, Fred. Thanks for coming.

[C]

Thanks for inviting me.

[B]

Yeah, I'm familiar with organoids, brain organoids and all that, but those are, you know, really simple objects. What are these mini brains that you guys. They're creating? What. What are they made of? How are they constructed?

[C]

Well, actually, no, the term mini brain is the. The term people are used for actually. Brain organoids. In the science committee, you would. You would talk about brain organoids or different types of organoids. Mini brain is not really realistic because it's very far from a brain.

[B]

Okay, so these are organoids then?

[C]

Yes.

[B]

Okay, so you'll take like one type of brain cell and you'll culture them and it'll make like a sphere or what kind of structures are they making?

[C]

Well, yes, most of the time look like it's like a sphere. Most of the time I would say, okay, if they are not like a sphere, it may be a problem, actually. Or it may be that you are differentiating in many different types of cells and then you have a guided differentiation, and this is also a topic of research is how different types of cells do you want to have in your organoids and how internally it is structured.

[B]

Okay, so are you keeping it simple with just one cell type and very little structure of any, or are you doing multiple cel.

[C]

We, we. We do both, actually. Okay. On. On one hand we do simple ones, and the other one we try to make advanced ones.

[B]

Is it. So what, so you're having them power computers? Why? What's special about a brain that would power a computer better, faster, with less energy? What's the benefit?

[C]

Well, when we say power, we means the computation. Okay. The power is. That's really a big power coming from the neurons. Okay. Of course. So. Well, there is. There are two things. Okay. Most obvious is the power efficiency. Okay. Your brain runs on 20 watts, but you have 100 billion of neurons, 10,000 synapses per neuron. If you wanted to make a simulation of this, that means basically run on a computer, traditional computer, your own brain will take like a small nuclear plant, you know.

[B]

So is it. So it's that complex just.

[C]

Yeah.

[B]

So if you were going to recreate this in silico, it would take enough power to power a nuclear power, it

[C]

would use enough power so that you need a power plant to power your silicon based simulation.

[B]

But somehow biology is doing it with 20 watts.

[C]

Yes, that's, that's the intriguing part. Actually, in biology, every neuron is actually a computer. Okay. And when you do AI using simulations of neurons, but in order to simulate, I don't know, 100 millions like when you use ChatGPT, for instance, you are going to have in memory 100 million artificial neurons. But when you want to update them, you are going to, to transfer all this data and compute it in one central computers or as many computers that you have. And that's always way, way less computers, even on GPU than the number of neurons. Which is not the case where the organic nervous system is using as many computers as number of neurons.

[B]

Also, our organic brain uses orders and orders and orders of magnitude more to little computers to establish the network. Is that what you're saying?

[C]

Well, yes. In addition, indeed, we have many, many ne connections in our brain. Okay, 100 billion is really impressive. But if you multiply now by 10,000 connections per neurons, it's something. Now there is a second reason why it's interesting is that, you know, we people think about AGI, you know, artificial general intelligence. Some of people think that maybe LLM is not going to be sufficient to reach AGI. And one approach, because there may be some missing something. Okay, one approach to this would have a hybrid approach to AGI when you have LLM on one side. So silicon based, I would say, for the reasoning through words, I would say, and the organic part, which is based on wetware and both being integrated together, they integrate actually very, very well. Digital world and organic world like wetware.

[B]

Right now with LLMs, it's the totality of, I mean, we don't get everybody, but it's like the totality of human thought on a subject. But the reasoning part, I would say, doesn't just only come from our experience, but we have. This is just me saying it, you know, I don't know, we somehow have a sophisticated, you know, apparatus that allows us to think. It doesn't seem like LLMs think, but they just explore the connections in the answer space of all the data that they've sucked up, but there's no like reasoning beyond it or on top of it. Is that right?

[C]

That's a complex question. Very complex. You know, of course you're right. Okay, Basically you are just doing interpolation in latent spaces. This is a general interpolator, artificial neurons. And that's it. This is. Okay, chatgpt, okay, in a nutshell, okay, very, very good. Are we way better than this? So for sure we are not manipulating words when we are reasoning. Maybe actually we don't maybe need words. Actually, it's not even sure that the brain was made for thinking. Okay. The brain is actually made for reacting and processing information, not thinking. So we don't feel that far superior to artificial neurons which are made to agglomerate words. I think these are two different ways to process information and both of which are very valid.

[B]

So. Oh, if we go back to the power consumption, by what factor less is power required needed to run a human brain versus doing it in silico? Is it like a million times less or what's the factor? Approximately?

[C]

Yeah, one million is. Okay. It's typically the factor that you can find on some publications like Professor Artung published on this. And now, you know, there is 1 million. This raw difference in power efficiency. Correct. However, if you implement this for real, you are going to run into a number of technicalities which are going to consume some of this energy efficiency. And I think at the end we'll be able to develop servers that will be maybe 100 to 1,000 times more energy efficient than the ones of Amazon Web Services, for instance.

[B]

Wow, what's happening? Haven't you been able to experimentally hook up a, you know, an organoid to, you know, to provide computation for some kind of chip? And has it worked? Like, what's happened?

[C]

Well, we do it on a daily basis. Okay. We have thousands of brain organoids in our lab and we put them on electrodes and we make them available for researchers across the world and they access it remotely in Python, so they interact with them and they work on programming it's and send stimulations, read their reaction, deliver neuromodulators and things like this entirely remotely.

[B]

Wow. Is it just so you pay for access to just one organoid or can you get. Do you have anyone, any of them that are daisy chained or tied to each other as well to make a little network?

[C]

Well, daisy chaining. Yeah, well, actually we did that. We did A number of experiments. This is called assembloids, actually. You take several assembly. Yeah, yeah. If you want to make it more complicated, it's easy actually. So for instance, we made organoids of three atom an hippocampus and cortical neurons. So three organoids. We put them with daisy chain, as we say, and then they will fuse together in a matter of days. Then you expect to have actually potentially the memory capabilities of the hippocampus, the capability to react to a reward of the striatum and the capability to compute of the cortical neurons.

[B]

So, okay, I don't know how well are the mini brains working for computation? Are they as efficient as if it was just all biological or are they a mixture of efficiency between in silico and biological?

[C]

So you don't want to use wetware computers for number crunching. Okay. Just don't do this. If you run Windows on a wetware computer, you're not going to be happy to be.

[B]

What are they using them for? Like, like are you using it for. Yeah. How are you using them then?

[C]

Well, first people are using it for research. Okay. It's. It's not in production like exploitation. Okay. Stage yet. Okay. But what is going to be the main target is going to be AI, Generative AI. Okay. Each time. Well, most of the AI today works with simulations of neurons. Okay. So instead of simulating, let's use the real ones. Okay. As simple as this. Very basic reasoning.

[B]

So. So something crazy. Has someone been able to like, what is the minimum number of organ needed to create an LLM powered by organoids?

[C]

So no, I can give you some order of magnitudes. Like if you want to simulate one neuron, one biological neurons, you would need several thousands of artificial neurons. Yeah. So you can take this the other way around actually. You would potentially use way less biological neurons than artificial neurons. Like thousands.

[B]

Oh, right, right. So saying like, like, like ChatGPT or Deep Seq or whatever.

[C]

Yeah.

[B]

Recreated and powered by organoids. And if so, how many would it take?

[C]

So this is really a bet. Okay. Based on a number of applications, I would bet that you need way less neurons. Okay. Biological neurons. And yes, I think it's absolutely possible to recreate a chatgpt. And that. That makes sense.

[B]

Barely a huge milestone, don't you think?

[C]

Yes, but it's a very meaningful milestone and it's a very conservative milestone to claim that ago to do information processing of information using biological neurons. Because nature has been doing this for a hundred millions of years.

[B]

Right, but. But you wouldn't Be able to explore the answer space. Normally, you know, with these organoids, you're. You're still using it as an LLM, not as like a regular biological brain. You're kind of using it for a different function. But it would be able to if. If it can mimic ChatGPT would be powered by again, these, these organoids. Have you guys made a calculation of how many organoids would be needed?

[C]

We what each organism has would say like 10,000 of neurons because these are mini organ brain organoids. Okay. In order to do servers, we will not use organoids anymore. We'll use big nervous tissues of metals of size.

[B]

Oh, okay. Because it's far more efficient than an organoid.

[C]

Yeah. It's going to be a totally different size.

[B]

Okay. I mean maybe this, I don't know if this is ridiculous or not, but could you take the brain, a part of the brain from an animal that has passed away and keep it alive and use that as, you know, a whole set of read to go neurons like a neuronal cortex to power computation? Your rat brain, you know, part of a rat brain.

[C]

Well, people have been doing works in electrophysiology, are doing exactly this for tens of years. Okay. Of studying primary cells of mammals, for instance. But this is absolutely not the direction that we are taking in our lab.

[B]

And I should have asked you before. I'm sorry, Go ahead, tell me what, what you're working on.

[C]

You're right. This is called primary cells. And that could be done also. Okay. But there is a problem with what you say is that you actually lot of animals and there is a more modern way to do this is actually to use stem cells. And then when you have stem cells, you can create millions of neurons in your lab.

[B]

Ah, yes. Okay.

[C]

Much more convenient, I can tell you.

[B]

This will be like a neuronal mass.

[C]

Yeah.

[B]

And that would be the organoids. It would be like a neuronoid or something like that.

[A]

Yeah, yeah.

[C]

It will be a big mass of nervous tissue. Yeah, ultimately, yes.

[B]

Have you made any of those yet?

[C]

No, no, the big ones. No, we. We made thousands of small ones of 1 millimeter of diameter. We still need to how to make them big. Big is easy, but big and living is more complicated to the blood supply.

[B]

Right?

[C]

Yes. The problem is vascularization. How do you breathe?

[B]

Yes. Oh, sorry. Go ahead. Yes, Okay, I understand. Gotcha. How do you bring blood supply so that nothing is more than a millimeter away, let's say for a blood supply or less than that, a micron away. Yes.

[C]

Yes. You're Right, Yeah, that's. That's the change.

[B]

Okay, so what's the major focus right now of your company? What are you guys trying to develop?

[C]

Well, there are different directions. First, we're using a lot of Agent Aki, the traditional ones, to interact the neurons. So basically you pile up some black boxes and. But agentic AI are able to. We give them API to interact directly with the neurons and read from the neurons. So they're able to perform things that we are not that fast and that smart to, you know, to modify the synaptic connections. You need to do a lot of things which are different from each algorithm. And Agent Ki are better than us, I think, for this. And it's not easy to figure out a universal algorithm. So so far. So that's a very revolutionary approach. And, and we.

[B]

Please, do you start with a structure and then allow the AGI or the GenIC AGI AI to reorder the structure of it to make it more efficient? Is any of that going on or once the structure set its statum.

[C]

No, no. If you talk about structure, you talk about connectivity between neurons. Okay. And when you change the connection between the neurons, you're not looking for efficiency at all. Okay. This is the list of your worries. What you want is the capability to. To learn something. Something that means rewiring the neurons. Actually, this is what you do.

[B]

They're giving like adjusting the weights in a, you know, in a neural network.

[C]

Yes, exactly this. Exactly this.

[B]

Yes. You're not changing the structure, but you're changing. By changing the weights, you're changing how the. The calculations in each layer occur. I understand. That makes sense.

[C]

Well, actually, in biological neural networks, you have actually no weight as such. You have synaptic connections and you have to change the synapses position in the network to change the weight.

[B]

Oh, wow. The algae is amazing. It's amazing how efficient and complex it is at the same time. It's unbelievable.

[C]

Yes, it is.

[B]

So, okay, so what would be like a dream result for your company in the next couple years? Is there one, one specific project that, you know, if it gets completed that you guys will be doing be really contributing like crazy?

[C]

Well, yeah, we have to master training of the alvolums. We have to scale up in order to. To become a serious competitor to Amazon Web Services. We are going to do it.

[B]

Okay, well, very good. Fred, where's the best way for people to keep tabs on what your companies are doing?

[C]

Oh, I think one good thing is to come on our Discord server. It's located directly from our website and people are just free to come here chat and we will answer to every question.

[B]

Okay, well very good. I know you know I started things late, but it still is a great short interview. So thank you so much. Fred, Appreciate you coming.

[C]

Thanks for your time and have a nice if you like this podcast, please

[B]

click the link in the description to subscribe and review us on itunes.

[A]

Thank you for listening to the Good Question podcast. Please email Support the GoodQuestion podcast.com if you have any referrals to great guests for us to interview. Visit thegoodquestionpodcast.com to hear more interviews. And please help us spread the word by rating and reviewing us on Apple podcast, podcasts, iTunes, Spotify, YouTube, or wherever you listen to this podcast.