4 ms·

The only problem with Elon's idea is that general AI does not exist and there is no theoretical framework at present for developing it. Even limited AI is mostl

by ackfoo 9y ago

The only problem with Elon's idea is that general AI does not exist and there is no theoretical framework at present for developing it. Even limited AI is mostly an over-hyped pile of crap. This is comparable to Elon's plan to retire on Mars. There is so much basic science to be done before it is feasible that it doesn't make sense to promulgate it as if it is an imminent reality.

pamqzl 9y ago
At least we know how to get to Mars, and roughly what new technologies we'd need for a colony. For AGI we have no clue even roughly how difficult it is. My own suspicion (and hope) is that we will eventually find we're able to build arbitrarily intelligent-at-some-task machines which avoid the unnecessary burden of having their own consciousness. I suspect consciousness is a nasty and inefficient evolutionary kludge rather than something actually necessary to perform specific tasks.
- observation 9y ago
  It is surely true we have no idea how difficult constructing an AGI is. It could require wholly new forms of hardware e.g. photon computation or parallelism similar to the brain's neuron might be a prerequisite. It is intriguing (I think) we know AGI is possible, I think we get there by emulations/ems if in no other way. Civilizations might rise and fall before it becomes feasible. I expect consciousness to be a lower level product, the development of language, reasoning, arithmetic, those are high level products. What's odd about our efforts in computer science is that we are plausibly going to be able to replace the higher functions without the lower ones such as mobility, consciousness. Moravec's Paradox is a kind of deep truth, perhaps more than we presently understand. The other inkling is that there is something Lovecraftian about the brain or intelligence in general. Consider how upsetting the atomic weapons were in the 20th, evolution theory in the 19th. I think it likely that venturing into the interior will reveal a truth mankind shall feel is at least as horrifying, probably more than giant bombs or genetic ancestry because it would register as being inescapable.

by ackfoo 9y ago

pamqzl 9y ago

At least we know how to get to Mars, and roughly what new technologies we'd need for a colony. For AGI we have no clue even roughly how difficult it is. My own suspicion (and hope) is that we will eventually find we're able to build arbitrarily intelligent-at-some-task machines which avoid the unnecessary burden of having their own consciousness. I suspect consciousness is a nasty and inefficient evolutionary kludge rather than something actually necessary to perform specific tasks.

observation 9y ago
It is surely true we have no idea how difficult constructing an AGI is. It could require wholly new forms of hardware e.g. photon computation or parallelism similar to the brain's neuron might be a prerequisite. It is intriguing (I think) we know AGI is possible, I think we get there by emulations/ems if in no other way. Civilizations might rise and fall before it becomes feasible. I expect consciousness to be a lower level product, the development of language, reasoning, arithmetic, those are high level products. What's odd about our efforts in computer science is that we are plausibly going to be able to replace the higher functions without the lower ones such as mobility, consciousness. Moravec's Paradox is a kind of deep truth, perhaps more than we presently understand. The other inkling is that there is something Lovecraftian about the brain or intelligence in general. Consider how upsetting the atomic weapons were in the 20th, evolution theory in the 19th. I think it likely that venturing into the interior will reveal a truth mankind shall feel is at least as horrifying, probably more than giant bombs or genetic ancestry because it would register as being inescapable.