This is so fascinating Gordon! I appreciate how you explain some deep tech ideas in a way I can make sense of it. The concept of wet tech or wetware has never crossed my mind, but as soon as you mentioned it, you had my attention. I'm working on a novel that includes robots. This article provided a whole new dimension for me to consider.
Thank you Lorraine! I’m glad you found it helpful. Wetware adds a very different layer to the usual robot story, because it raises the possibility that the “machine” is not entirely a machine anymore. That opens up a lot of interesting questions for fiction as well as science.
Great overview here. Besides Cortical Labs, what are the other startups or lab-based efforts you have found that are doing interesting commercial-oriented work here?
That’s a good question, and the short answer is that there still are not many companies in this space. Cortical Labs is the most visible one right now, with its CL1 biological computer and Cortical Cloud platform. FinalSpark is the other company worth watching. It offers a remote access Neuroplatform for experiments on brain organoids.
Most of the other serious work is still happening in academic labs. UC Santa Cruz has published work on goal directed learning in cortical organoids, and Johns Hopkins recently received major funding for its DROID platform, which combines brain organoids, sensors, and AI analysis. So my sense is that the field is real, but still early: a small number of actual companies, and a larger group of labs building the tools that could later become companies.
Thermodynamic Computing has been investigated for years. There have been some breakthrough in recent years where they proposed the new hardware for computation in companies like extropic and normal computing
That’s a great point. Thermodynamic computing is another serious attempt to get past the energy limits of conventional silicon, and it is fascinating because it may capture some of biology’s efficiency without using living tissue. What makes organoid intelligence different is that it does not just imitate biology’s dynamics, it tries to compute with biology itself.
Interestingly, I'm reading Jack Carr's "Red Sky Mourning", a fictional thriller. He mentions how the quantum computer in the story used to drive AI for the military will someday consume too much power. I'm well past that chapter, but I believe that he referred to a possible solution being "DNA Computing".
Having worked in the datacenter space for three decades, power requirements keep getting more dense. My first datacenter averaged 3 KVA per cabinet. Newer datacenters are exceeding 60 KVA per cabinet.
Increased power requires increased heat dissipation---Watts in=BTUs out. Air cooling with hot aisles/cold aisles no longer works well enough. The newer datacenters are having glycol-based liquid cooling circuits included to better move heat from server chassis.
Also on cooling and power, Dell had conducted a study about 10 or 15 years ago and determined that for every watt of power sent to servers required another seven watts of power toward cooling, power conditioning, etc.
So, having some kind of "wet" computing technology which reduces the power and cooling requirements sounds like it could be a gold mine, if workable.
That’s a great point, and your datacenter perspective gets right to the heart of why this matters. The power wall in AI is not an abstract problem anymore. It is showing up in rack density, cooling architecture, and the sheer infrastructure burden needed to keep scaling silicon. Once intelligence starts looking like a thermodynamics problem, biology becomes much more interesting.
I also like that you brought up DNA computing. It sits in the same broad family of biological computing ideas, but organoid intelligence is a bit different. DNA computing is usually about encoding or processing information chemically, whereas organoid intelligence tries to exploit the adaptive dynamics of living neurons themselves. In other words, the appeal is not only lower power use, but a different kind of computation.
That said, “wet” computing would not be free of engineering headaches. You trade some of the heat and electrical burden for sterility, nutrient delivery, monitoring, and biological variability. Still, if these systems prove useful for the right kinds of tasks, especially ones involving adaptation or pattern learning, I agree they could become enormously valuable.
Thinking about the human brain, it's got a great amount of computational capacity, it's low wattage, low BTU, but it's prone to errors. Its data durability is horrible! Look at how we forget little details. I walk into a room sometimes and wonder, "Why did I come here?"
Silicon-based computing is far less prone to errors, but it's computational capacity is limited. It's high wattage and high BTU.
Making some sort of wet computing, can the data durability be improved?
That is exactly the right question. The brain is extraordinarily efficient, but it does not store information the way a hard drive does. Biological memory is distributed, plastic, and noisy. In many cases, forgetting is not a failure so much as a feature, because it prevents overload and allows the system to stay adaptive rather than preserving every detail with bit-level fidelity.
So I suspect wet computing will not win by replacing silicon storage. It will win, if it wins, as a different kind of processor. You would still want silicon, or some other conventional medium, for durable long term memory, exact records, and error correction. The wetware would be the adaptive layer: pattern recognition, closed loop learning, novelty handling, and tasks where flexibility matters more than perfect recall. In that model, the durability problem is improved not by making biology act like a hard drive, but by pairing it with hardware that already does that well.
This is so fascinating Gordon! I appreciate how you explain some deep tech ideas in a way I can make sense of it. The concept of wet tech or wetware has never crossed my mind, but as soon as you mentioned it, you had my attention. I'm working on a novel that includes robots. This article provided a whole new dimension for me to consider.
Thank you Lorraine! I’m glad you found it helpful. Wetware adds a very different layer to the usual robot story, because it raises the possibility that the “machine” is not entirely a machine anymore. That opens up a lot of interesting questions for fiction as well as science.
Great overview here. Besides Cortical Labs, what are the other startups or lab-based efforts you have found that are doing interesting commercial-oriented work here?
That’s a good question, and the short answer is that there still are not many companies in this space. Cortical Labs is the most visible one right now, with its CL1 biological computer and Cortical Cloud platform. FinalSpark is the other company worth watching. It offers a remote access Neuroplatform for experiments on brain organoids.
Most of the other serious work is still happening in academic labs. UC Santa Cruz has published work on goal directed learning in cortical organoids, and Johns Hopkins recently received major funding for its DROID platform, which combines brain organoids, sensors, and AI analysis. So my sense is that the field is real, but still early: a small number of actual companies, and a larger group of labs building the tools that could later become companies.
Thermodynamic Computing has been investigated for years. There have been some breakthrough in recent years where they proposed the new hardware for computation in companies like extropic and normal computing
That’s a great point. Thermodynamic computing is another serious attempt to get past the energy limits of conventional silicon, and it is fascinating because it may capture some of biology’s efficiency without using living tissue. What makes organoid intelligence different is that it does not just imitate biology’s dynamics, it tries to compute with biology itself.
Interestingly, I'm reading Jack Carr's "Red Sky Mourning", a fictional thriller. He mentions how the quantum computer in the story used to drive AI for the military will someday consume too much power. I'm well past that chapter, but I believe that he referred to a possible solution being "DNA Computing".
Having worked in the datacenter space for three decades, power requirements keep getting more dense. My first datacenter averaged 3 KVA per cabinet. Newer datacenters are exceeding 60 KVA per cabinet.
Increased power requires increased heat dissipation---Watts in=BTUs out. Air cooling with hot aisles/cold aisles no longer works well enough. The newer datacenters are having glycol-based liquid cooling circuits included to better move heat from server chassis.
Also on cooling and power, Dell had conducted a study about 10 or 15 years ago and determined that for every watt of power sent to servers required another seven watts of power toward cooling, power conditioning, etc.
So, having some kind of "wet" computing technology which reduces the power and cooling requirements sounds like it could be a gold mine, if workable.
That’s a great point, and your datacenter perspective gets right to the heart of why this matters. The power wall in AI is not an abstract problem anymore. It is showing up in rack density, cooling architecture, and the sheer infrastructure burden needed to keep scaling silicon. Once intelligence starts looking like a thermodynamics problem, biology becomes much more interesting.
I also like that you brought up DNA computing. It sits in the same broad family of biological computing ideas, but organoid intelligence is a bit different. DNA computing is usually about encoding or processing information chemically, whereas organoid intelligence tries to exploit the adaptive dynamics of living neurons themselves. In other words, the appeal is not only lower power use, but a different kind of computation.
That said, “wet” computing would not be free of engineering headaches. You trade some of the heat and electrical burden for sterility, nutrient delivery, monitoring, and biological variability. Still, if these systems prove useful for the right kinds of tasks, especially ones involving adaptation or pattern learning, I agree they could become enormously valuable.
Thinking about the human brain, it's got a great amount of computational capacity, it's low wattage, low BTU, but it's prone to errors. Its data durability is horrible! Look at how we forget little details. I walk into a room sometimes and wonder, "Why did I come here?"
Silicon-based computing is far less prone to errors, but it's computational capacity is limited. It's high wattage and high BTU.
Making some sort of wet computing, can the data durability be improved?
That is exactly the right question. The brain is extraordinarily efficient, but it does not store information the way a hard drive does. Biological memory is distributed, plastic, and noisy. In many cases, forgetting is not a failure so much as a feature, because it prevents overload and allows the system to stay adaptive rather than preserving every detail with bit-level fidelity.
So I suspect wet computing will not win by replacing silicon storage. It will win, if it wins, as a different kind of processor. You would still want silicon, or some other conventional medium, for durable long term memory, exact records, and error correction. The wetware would be the adaptive layer: pattern recognition, closed loop learning, novelty handling, and tasks where flexibility matters more than perfect recall. In that model, the durability problem is improved not by making biology act like a hard drive, but by pairing it with hardware that already does that well.