Neurogrids

[Kai]

Oh great, mechanical monkeys.

[Reginald Ret]

i propose using both types of computers in the transition stage, that'll give 'em enough time to work out the making mistakes bit.

[Triple Zero]

Actually there are enough examples where it is perfectly fine to make efficiency / accuracy trade-offs like this.

In fact, this would be pretty fucking useful for databases. I've read some stuff about neural models for databases, and if you design them that way, they are able to deduce missing pieces of data and discover patterns and trends in the data as primitive operations like you'd normally do an exact query on a traditional DB. it's not always what you want, but there's quite some situations in which this would actually be better than a traditional DB.

it's kinda scary too, though. cause this is getting pretty close to those positronic brain things in Asimov's novels. being very complex, computer-designed networks of computer stuff that doesn't always work entirely right and the very worst thing is that it is way too complex for us ever to truly understand what is going on. Yes, kind of like the human brain, or many other processes in biology.

problem is, what if it works? what if it happens to be more efficient than biology?

you could argue that that's not gonna happen cause evolution has had millions of years to polish and perfect their biological processors, and we would be hard pressed to outdo nature on this field.
but, you don't know that. these bio processors were evolved for fitness, at every step along the way. which means they could only move through the ridges of the local maxima. if you add some human "intelligent design" into that mix, who knows where you can carry it.
I mean look at the regular computer,. such a thing wasnt gonna evolve by itself either.

[Rococo Modem Basilisk]

Oh great, mechanical monkeys.

I figure that's the goal.