It would probably resemble something more you'd expect in the field of neuroscience, where you have to probe and tune artificial neural firing through well-timed spikes. The chip architecture would resemble loose similarities to biological neural networks. So any code used would be on a conventional co-processor to initialize the AI chip through a sequence of spiking patterns. It could be extremely simple instructions like: Assigning where, when and how much of a excite/inhibit spike should be delivered.
[–] 16422718? ago
The first real functional AI will be written primarily on the hardware, so maybe not ASM, but very close to ASM.
[–] 16422721? ago
It won't be written, it will be evolved (using evolutionary algorithms).
[–] 16423526? ago
>>12729474
It would probably resemble something more you'd expect in the field of neuroscience, where you have to probe and tune artificial neural firing through well-timed spikes. The chip architecture would resemble loose similarities to biological neural networks. So any code used would be on a conventional co-processor to initialize the AI chip through a sequence of spiking patterns. It could be extremely simple instructions like: Assigning where, when and how much of a excite/inhibit spike should be delivered.