3 ms·

> why would the ai be millions of generations smarter? Pure uneducated speculation on my part. 1) In order to emulate a Super Nintendo you need hardware at le

by columbo 13y ago

> why would the ai be millions of generations smarter? Pure uneducated speculation on my part. 1) In order to emulate a Super Nintendo you need hardware at least 20x faster than the hardware needed to run the Super Nintendo natively. In order to emulate a human brain you'd need something with much more potential... if that potential exists why waste cycles emulating? If it isn't emulated then it is something else entirely. 2) We choose to rely on robots to explore the floor of the ocean with the occasional human being showing up just to make a documentary. We'll have cars driving us around, we use video for everything, we'll rely more and more on stuff brought to us instead of us going to various locations. 3) Even if we start with human intelligence the ability to remove unnecessary components will create something that is no longer human. I don't believe the needs/wants of an ai program or modified human construct will behave remotely human. Take a real basic element of human/animal behavior; survival. I don't think AI would find it necessary to 'defend itself' because it is a logical next-step for any advancing species. If it is destroyed today it'll come back in a ten, a hundred, or a million years. AI (if it is possible) is as inevitable as the printing press, lightbulb and stone mill. Of course, ymmv, I mean we may as well debate how many noodles the Flying Spagetti Monster has, there's only 1% science and 99% science fiction at this time.

csense 13y ago
> you need hardware at least 20x faster than the hardware needed to run the Super Nintendo natively Only if the emulating hardware is general purpose. Yes, simulating a human brain on a cluster of Intel i7's or AMD GPU's, or their future equivalents, will probably require an enormous hardware investment. But once you have one working emulated brain on general-purpose hardware, you can produce specialized hardware. E.g. if we knew that Super Nintendo held the secrets to immortality, we would soon make updated versions of all its chips with today's technology, and you could probably fit dozens or hundreds of SNES cores on a single modern chip with the decades of improvement in manufacturing processes. Likewise, once we make an initial discovery of how to emulate human brains on general-purpose computers, the things we find out from that implementation will start to pave the way for special-purpose computers that can more cost-effectively execute the brain emulation application. Also, it's entirely possible that you can emulate human brains, but you don't know how to create more potential. E.g., if we can simulate the physics and chemistry of human brains well enough to replicate cognitive processes, but our best experts are simply stumped at figuring out exactly which essential features of the physical/chemical processes translate into intelligence. And if early experiments at tweaking model parameters show that the human brain has evolved into some local maximum, where any tweak produces a less powerful intelligence or one that's defective in some way, and after this is observed over many experiments, it might then become taboo to tweak the model for ethical reasons, making the situation stable over the long term. In other words, we might be "forced" to "waste" cycles if we can't figure out which cycles we're wasting, and the experiments that might tell us would entail sacrificing too many artificial humans to be ethically acceptable. This state of affairs isn't necessarily likely, but it's certainly possible.

by columbo 13y ago

csense 13y ago

> you need hardware at least 20x faster than the hardware needed to run the Super Nintendo natively Only if the emulating hardware is general purpose. Yes, simulating a human brain on a cluster of Intel i7's or AMD GPU's, or their future equivalents, will probably require an enormous hardware investment. But once you have one working emulated brain on general-purpose hardware, you can produce specialized hardware. E.g. if we knew that Super Nintendo held the secrets to immortality, we would soon make updated versions of all its chips with today's technology, and you could probably fit dozens or hundreds of SNES cores on a single modern chip with the decades of improvement in manufacturing processes. Likewise, once we make an initial discovery of how to emulate human brains on general-purpose computers, the things we find out from that implementation will start to pave the way for special-purpose computers that can more cost-effectively execute the brain emulation application. Also, it's entirely possible that you can emulate human brains, but you don't know how to create more potential. E.g., if we can simulate the physics and chemistry of human brains well enough to replicate cognitive processes, but our best experts are simply stumped at figuring out exactly which essential features of the physical/chemical processes translate into intelligence. And if early experiments at tweaking model parameters show that the human brain has evolved into some local maximum, where any tweak produces a less powerful intelligence or one that's defective in some way, and after this is observed over many experiments, it might then become taboo to tweak the model for ethical reasons, making the situation stable over the long term. In other words, we might be "forced" to "waste" cycles if we can't figure out which cycles we're wasting, and the experiments that might tell us would entail sacrificing too many artificial humans to be ethically acceptable. This state of affairs isn't necessarily likely, but it's certainly possible.