AIs using lab-grown "brains" are getting closer
A new paper in the journal Nature Electronics demonstrates that brain organoids can be used as simple organic neural nets, increasing the likelihood of organic AI's in the future
If you thought that 2023 was a far-out year for AI, it’s about to get weirder!
Back in May I wrote about a concept paper establishing the field of “organoid intelligence” — using lab-grown brain organoids as the substrate for increasingly powerful artificial intelligence systems.
A new paper in the journal Nature Electronics has now shown the use of brain organoids in constructing simple neural nets that can be trained to carry out basic functions.
Brain organoids are self-assembling clusters of brain cells grown in the lab that, together, begin to exhibit structures and properties that mimic the human brain. They’re especially interesting as a substrate for AI in that they combine data storage with data processing at a granular level (overcoming what’s known as the von Neumann bottleneck) while demonstrating a remarkable ability to learn and compute while using a fraction of the energy required by a digital AI systems.
And while progress is being made on neuromorphic chips that mimic biological neurons, evolved biological brains still reign supreme in the world of intelligence.
It’s not surprising therefore that the possibility of growing brain-like structures in the lab and hooking them up to silicon-based AI systems are beginning to attract attention.
Reflecting this, the authors of the Nature Electronics paper “Brain organoid reservoir computing for artificial intelligence” connected a brain organoid grown from human pluripotent stem cells to a high density array of electrodes, and started to explore what they could train it to do.
In the paper, they approach two relatively simple tasks: recognizing aspects of speech from a specific individual, and predicting the behavior of a nonlinear and chaotic dynamic system.
In each case, the hooked-up brain organoid demonstrated an ability to be trained and, following this, to carry out the assigned tasks when presented with appropriate inputs.
In the grand scheme of things, these capabilities on their own are not earth shattering. But the novelty of a compute substrate that’s built on human brain cells does open up some intriguing new possibilities.
Here, the authors speculate that brain organoid-based AIs may have the “necessary complexity and diversity to mimic a human brain, which could inspire the development of more sophisticated and human-like AI systems.”
And coining the term “Brainoware” technology, they’re optimistic that lab grown brain organoids could further extend what is possible with AI.
This aligns with the concept paper on Organoid Intelligence published in May, and helps to translate these ideas from concept to practice.
Of course, there’s still a long way to go before we’re growing the next generation of AIs. But step by step, we’re getting there.
That said, if “Brianoware” or other forms of organoid intelligence do take off, this does put a slightly more challenging twist on the question of what happens if these bio-AI’s develop consciousness. And if they did, how we’d know …