I was not just astounded but also amazed at how innovations can go this far. It is just funny that the brains behind all these technological innovations will also use brain cells to power these innovations.
The CEO of Cortical Labs Hon Weng Chong is building biological computers that according to him has the capability to be better than Artificial Intelligence (AI) – smarter and more efficient – at the same time slash the enormous energy costs in training AI. This is by combining the learning ability of human brain cells and the processing power of silicon chips (source: Forbes Asia: Head First by Zinnia Lee July 2023). Chong was recently successful in using a brain cell-powered computer chip “Dish Brain” that has learned to play Atari’s Pong game while using energy so small that is estimated to just be equal to the energy being used by a calculator. This ability of the brain cells to play Pong can be used as a benchmark for optimizing the models and discovering more ways to use brains to work as computers although it can also be tricky and not easy.
Machine learning like teaching AIs, requires electronic approaches that consume ever-growing computing power and high energy demands. Furthermore, the recent trend of “neuromorphic computing” which aims to mimic electronic human neural activity has inherent limitations of conventional electronics.
Using human brain cells however, is capable with lesser effort in performing the above functions and requires a lot less energy or power demands. And speaking of energy usage, training AI GPT-3 (I use GPT-3 oftentimes in my research) took 1.287-gigawatt hours or about as much electricity as 120 US homes would consume in a year and carbon emissions of 502 metric tons of carbon equivalent to driving 112 gasoline-powered cars for a year, according to research in an article “AIs Growing Carbon Footprint” dated June 9, 2023 (source: Carbon Emissions And Large Neural Network Training).
On the other hand, “organoid intelligence” or using brain cells or brain organoids grown in cell culture that share some of the functions and structure as brains use a lot less energy and have smaller carbon footprints and are not as vulnerable to traditional electronic risks.
A classic example is the power consumption of the new supercomputer – it consumes an average of 21 megawatts. The human brain, on the other hand, operates at the same 1 exaFlop and consumes only 20 watts! Wow – 20 megawatts against 20 watts only! It means that our brain operates at the estimated same 1 exaFlop but at a 10 raised to the power of 6 – fold better power efficiency compared to modern machines although performing quite different tasks(source:frontiersin.org).
That is the reason why there is a race towards the better use of organoid intelligence to not only function better than artificial intelligence but a lot more energy efficient. It can also be used to power the learning process of artificial intelligence at a lot less energy costs. And training a biological computer takes a lot fewer hours than training an AI. It takes 10 raised to the power of 10 times more energy for AI to learn a task compared to human brain cells.
But like you, I am curious how can the brain cells be utilized to power or even function as biological computing and perhaps even be faster and more efficient and more powerful than silicon-based computing and AI and using a lot less energy. This is what I like about being a writer - it forced you to research something which caught your curiosity. In the past decades, there has already been a revolution of brain cell cultures – starting as traditional monolayer cultures to more organ-like, organized 3D cultures like brain organoids.
But where do they get these cells? These are generated from embryonic stem cells which can draw ethical problems or derived from skin samples which can be less ethically problematic. But culture times can sometimes be prolonged which can exceed 1 year (source: frontiersin.org dated
February 28, 2023).
Using brain cells means it uses 3D cultures of human brain cells (brain organoids) and brain-machine interface technologies. The research on this could also help in understanding brain development, learning, and memory possibly helping treatments for neurological disorders such as dementia. Now that is good news for those who are worried about dementia in old age! But what are the downsides?
Most research however on organoid intelligence is still in the infancy stage although the concept of brain-machine interfacing emerged around five decades ago. While AI aims to make machines perform more like human brains, the current research objective is to explore different ways in which a 3D brain cell culture can be made more like how a machine or computer works.
We are busy doing our everyday work and living our personal lives day-by-day and we will just wake up one day that our brains are now used to run computers at a higher power and efficiency than how a silicone-based machine is. How can we be sure that some bright humans in the future who will discover attaching human brains to a computer will use it for the good of mankind rather than to use its power for their own selfish benefits?
Organoid intelligence may drastically reduce energy bills but is it worth the moral risks?
(Wilma Miranda is a Managing Partner of Inventor, Miranda & Associates, CPAs, Chair of the Ethics Committee of FINEX and member of the Board of Directors of KPS Outsourcing, Inc. The views expressed herein do not necessarily reflect the opinion of these institutions.)