Biocomputing is being hailed as the future of artificial intelligence. But the technology depends on growing and discarding human brain tissue.
www.theblaze.com
Sound pretty dark to me, enslaving the stem cells of dead unborn babies to become part of AI?
Naturally, the Left will love this. Not only will there be an even larger market for abortion, but they can also build their "god" AI in the process to worship and follow as they use the knowledge to empower themselves.
Knowledge without wisdom, that is nirvana for the Left.
Yay.
Don't be worried, it's a passing fad.
It started back when neural networks became a thing, about 15 years ago. There's certain kinds of computations that AI simply can't do, that real neurons can.
So someone figured out how to culture cells on a chip, and the technology had/has very limited (almost no) commercial success.
Today, photonics is the thing. The bridge between AI and quantum computing is where it's at. Just today, a Chinese team announced a metasurface capable of entangling large numbers of photons, either one at a time or all at once
For actual brain research, neurons on a chip are useless. We already have technology that far surpasses it in capability.
The only part of brain computation that AI can't yet do is the non-algebraic part, the part that involves oscillations and phase coding. Very few people understand how this actually works (I do, if you'd like to know about it I'll tell you, and show you). None of that is present in these neurons-on-a-chip, they're strictly algebraic (nonlinear, but otherwise not very smart).
Neurons by themselves are kind of dumb, interesting but not very capable. You have to get them in large populations to do brainy stuff. You can program a neuron genetically (for example, to oscillate) but it's pretty hard, it takes weeks of biochemistry whereas you could get on Google AI and do the same thing algebraically with TensorFlow in minutes.
Technically the main thing the live milieu brings to the table is "active learning". AI works on an error-correcting scheme called "back propagation" which is highly non-biological but works real good with digital computers and GPU's. Competing and more brain-like methods have been around since the 80's, they're just not as fast in the digital domain because you have to solve differential equations by numerical approximation.
Real neurons on a chip have no future, commercial or otherwise. Photonic neural networks can already classify objects in visual space in less than 500 picoseconds, real neurons can't touch that. Real neurons depend on water diffusion, which is real slow compared to the speed of light.
What is needed is a software language to program quantum computers at a high level. A company called D-Wave almost has one, but it's architecture specific and therefore of limited use. The demand is great though, someone will invent one soon.