3 ms·

This isn't a solution and is normally considered very early on in any discussion of the paradox. A replicating machine intelligence can presumably be very pati

This isn't a solution and is normally considered very early on in any discussion of the paradox. A replicating machine intelligence can presumably be very patient, and an early civilisation evolved into that, or using it to seed colonies or just investigate could have populated the galaxy many times over by now.

simonh 7y ago
It’s considered yes, but IMHO it’s often dismissed out of hand without any actual credible analysis. For example, I suspect that any self replicating machine would need to be run by an AI. So it would have to be able to manufacture, from base materials, a computer capable of running an AI and a duplicate of all the industrial mechanisms needed to do that again, from the silicon chips up. That, plus the infrastructure to construct metal foundries, chemical works, propulsion systems, power reactors, the list goes on and on. It would have to be absolutely enormous, and sending it to another solar system is basically throwing it away from a practical point of view. It such a huge undertaking I doubt think anyone has ever seriously done any analysis of hat would be involved. That’s not a good sign, when a task so many people seem to write off as trivial, is so hard that nobody has even managed to do an overall actual engineering assessment or draft plan.
- amelius 7y ago
  Perhaps you should read Kurzweil: https://en.wikipedia.org/wiki/The_Singularity_Is_Near https://en.wikipedia.org/wiki/The_Singularity_Is_Near
  - simonh 7y ago
    I have, many of his articles and interviews, and I think he’s way off the mark. Technological advancements follow S curves, not eternal exponentials. Computer technology advancement is already flattening out and no new fundamental advances are anywhere near taking over from the current technical architectures. Aside from which, his work doesn’t actually address any of the challenges I described. He tries to wave it away with the magic singularity wand, but doesn’t provide any actual technical roadmap to AI. I do believe we’ll get there eventually, I’m not an AI denier, but I think I have a realistic sense of the limitations we face given the actual physics of the universe, and the limitations of the materials it’s composed of.
- mellosouls 7y ago
  Possibly, but also possible you are prematurely overestimating the constraints and limiting the abilities and knowledge of a civilisation presumably hundreds of years ahead of us at a minimum. Countering your other reply, I acknowledge physics limitations like the maximum speed are unlikely to change, but see no reason for similar constraints on progress in self-replicating intelligence (apart from our current understanding of technologies involved like AI, which IMO is vastly overhyped). By "machine intelligence", I meant AI (a term I don't like), but see no reason why it should be constrained to metal chips and boxes as it might be using our current technology.
- abecedarius 7y ago
  A self-replicating machine can fit in a cubic micron. For a machine that can build a dish antenna out of asteroid material and receive beamed instructions to build more complex machines, the seed probably would have to be bigger than a bacterium, but it seems disproportionate to imagine the complexity has to be greater than the seed of a tree. On the one hand, it has to grow in a very different environment, but on the other it's an environment without voracious competitors looking to eat or outgrow it.
  - simonh 7y ago
    The context is interstellar probes to colonise the galaxy, that requires a self replicating universal constructor. Otherwise you just get a grey goo, or on Ian M. Banks terms a self replicating hegemonic swarm.
    - abecedarius 7y ago
      "First build a radio receiver" means you don't have to pack anything like your whole civilization into the seed, unless you're in some kind of lifeboat emergency like a war or supernova. Biology is an existence proof of universal-enough constructors (and a radio is simple compared to a tree). These existing constructors may seem absurdly hyper-advanced compared to current human technology, but I think that's a prejudice. Even current computers are hyper-advanced over the ones of just a lifetime ago. It's hard to plausibly avoid that level of progress over a few lifetimes more, short of somehow stopping progress cold.

simonh 7y ago

It’s considered yes, but IMHO it’s often dismissed out of hand without any actual credible analysis. For example, I suspect that any self replicating machine would need to be run by an AI. So it would have to be able to manufacture, from base materials, a computer capable of running an AI and a duplicate of all the industrial mechanisms needed to do that again, from the silicon chips up. That, plus the infrastructure to construct metal foundries, chemical works, propulsion systems, power reactors, the list goes on and on. It would have to be absolutely enormous, and sending it to another solar system is basically throwing it away from a practical point of view. It such a huge undertaking I doubt think anyone has ever seriously done any analysis of hat would be involved. That’s not a good sign, when a task so many people seem to write off as trivial, is so hard that nobody has even managed to do an overall actual engineering assessment or draft plan.

amelius 7y ago
Perhaps you should read Kurzweil: https://en.wikipedia.org/wiki/The_Singularity_Is_Near https://en.wikipedia.org/wiki/The_Singularity_Is_Near
- simonh 7y ago
  I have, many of his articles and interviews, and I think he’s way off the mark. Technological advancements follow S curves, not eternal exponentials. Computer technology advancement is already flattening out and no new fundamental advances are anywhere near taking over from the current technical architectures. Aside from which, his work doesn’t actually address any of the challenges I described. He tries to wave it away with the magic singularity wand, but doesn’t provide any actual technical roadmap to AI. I do believe we’ll get there eventually, I’m not an AI denier, but I think I have a realistic sense of the limitations we face given the actual physics of the universe, and the limitations of the materials it’s composed of.
mellosouls 7y ago
Possibly, but also possible you are prematurely overestimating the constraints and limiting the abilities and knowledge of a civilisation presumably hundreds of years ahead of us at a minimum. Countering your other reply, I acknowledge physics limitations like the maximum speed are unlikely to change, but see no reason for similar constraints on progress in self-replicating intelligence (apart from our current understanding of technologies involved like AI, which IMO is vastly overhyped). By "machine intelligence", I meant AI (a term I don't like), but see no reason why it should be constrained to metal chips and boxes as it might be using our current technology.
abecedarius 7y ago
A self-replicating machine can fit in a cubic micron. For a machine that can build a dish antenna out of asteroid material and receive beamed instructions to build more complex machines, the seed probably would have to be bigger than a bacterium, but it seems disproportionate to imagine the complexity has to be greater than the seed of a tree. On the one hand, it has to grow in a very different environment, but on the other it's an environment without voracious competitors looking to eat or outgrow it.
- simonh 7y ago
  The context is interstellar probes to colonise the galaxy, that requires a self replicating universal constructor. Otherwise you just get a grey goo, or on Ian M. Banks terms a self replicating hegemonic swarm.
  - abecedarius 7y ago
    "First build a radio receiver" means you don't have to pack anything like your whole civilization into the seed, unless you're in some kind of lifeboat emergency like a war or supernova. Biology is an existence proof of universal-enough constructors (and a radio is simple compared to a tree). These existing constructors may seem absurdly hyper-advanced compared to current human technology, but I think that's a prejudice. Even current computers are hyper-advanced over the ones of just a lifetime ago. It's hard to plausibly avoid that level of progress over a few lifetimes more, short of somehow stopping progress cold.