What’s your success criterion? Do you mean a human mind that the unuploaded copy will accept as a successful upload? Or that the relatives will accept? Or that some panel of expert judges will accept? In the latter cases, will it have to be unanimous?
Some people with particularly detailed Facebook timelines can conceivably be emulated well enough to fool the very gullible without any uploading taking place at all. Very senile people would also be easy to emulate. Babies would be easier than people with complex memories. Very rational people would be easier than those with idiosyncratic patterns of reasoning. People who do work that is hard to characterize (like architecture) would be easier to emulate than those who do work we find easy to characterize (like fiction writing). And so on.
I imagine, on the one hand, a brain scan and emulation system that convinces a couple of aging relatives that granny is now in the computer. And on the other hand, a system that allows a team of expert scientists to keep working together after the demise of one of them. Where on this spectrum is what you mean?
Because I wouldn’t be surprised if the former took a million times less memory and computational power than the latter.
People who do work that is hard to characterize (like architecture) would be easier to emulate than those who do work we find easy to characterize (like fiction writing).
I bet some architects are doing fairly routine work. I’m sure that some writers do work which is hard to characterize. Stephen King does gore by the yard, but every now and then he writes a story which is different from his usual, and that’s the part which would be hard for an em to get right.
Considering Scott Alexander, I don’t think it’s adequate to contrast very rational people against people with idiosyncratic patterns of reasoning.
More generally, there are computer programs which compose music by using patterns from major composers, and result is “sounds like what Bach would write on an off day”. The hard challenge is to get the next “Jesu, Joy of Man’s Desiring”.
I think we are committing some fallacy here. The same fallacy that leads to people to judge things simple once they are understood.
The real complexity of real life lies in interdependencies that are hard to capture and make precise. That doesn’t mean that anything humans do can’t be made precise and understood in principle. It just means that the things we understand seem simple.
How about, “Able to be employed at the same jobs as the original, and (if run at realtime speeds) able to perform as well as the original on any task not involving physical labour”, with ‘same jobs’ including anything from light office work to original academic research?
(I’m hoping to learn something in this thread which I can apply to economics once ems exist, and the above seems to closely correspond with the economic impact of the existence of an em.)
When very small shell scripts can be used to replace the researchers who are coming up with ways to keep computer improvements resembling Moore’s Law coming, then that’ll be a bit more relevant. :)
“Computer” used to be a job for humans. As Marc Andreessen pointed out long time ago, software is eating the world—note the present tense, no need to wait for AIs or uploads.
And some jobs are more amenable to replacement than others. (Eg, “Manna”, by Marshall Brain, at http://marshallbrain.com/manna1.htm .) It seems safe to say that being able to create ems would be a significant step in replacing jobs currently held by biological humans; but there are all sorts of details involved which change the economic equations, such as how much RAM an em requires, and when the first em comes online. I’m afraid that your statement doesn’t seem to offer anything that I can use to improve my current estimates on any of these matters.
I am still not going to be particularly useful :-/
With respect to uploads/ems my position is hardcore Knightian uncertainty: not only I don’t have any estimates of the timing, I will disbelieve any estimates other people produce as well.
In fact, I don’t know if one can generate an upload by brain scanning at all. I certainly don’t think it’s an inevitability only delayed by the need to develop appropriate tech.
What’s your success criterion? Do you mean a human mind that the unuploaded copy will accept as a successful upload? Or that the relatives will accept? Or that some panel of expert judges will accept? In the latter cases, will it have to be unanimous?
Some people with particularly detailed Facebook timelines can conceivably be emulated well enough to fool the very gullible without any uploading taking place at all. Very senile people would also be easy to emulate. Babies would be easier than people with complex memories. Very rational people would be easier than those with idiosyncratic patterns of reasoning. People who do work that is hard to characterize (like architecture) would be easier to emulate than those who do work we find easy to characterize (like fiction writing). And so on.
I imagine, on the one hand, a brain scan and emulation system that convinces a couple of aging relatives that granny is now in the computer. And on the other hand, a system that allows a team of expert scientists to keep working together after the demise of one of them. Where on this spectrum is what you mean?
Because I wouldn’t be surprised if the former took a million times less memory and computational power than the latter.
I bet some architects are doing fairly routine work. I’m sure that some writers do work which is hard to characterize. Stephen King does gore by the yard, but every now and then he writes a story which is different from his usual, and that’s the part which would be hard for an em to get right.
Considering Scott Alexander, I don’t think it’s adequate to contrast very rational people against people with idiosyncratic patterns of reasoning.
More generally, there are computer programs which compose music by using patterns from major composers, and result is “sounds like what Bach would write on an off day”. The hard challenge is to get the next “Jesu, Joy of Man’s Desiring”.
I think we are committing some fallacy here. The same fallacy that leads to people to judge things simple once they are understood.
The real complexity of real life lies in interdependencies that are hard to capture and make precise. That doesn’t mean that anything humans do can’t be made precise and understood in principle. It just means that the things we understand seem simple.
How about, “Able to be employed at the same jobs as the original, and (if run at realtime speeds) able to perform as well as the original on any task not involving physical labour”, with ‘same jobs’ including anything from light office work to original academic research?
(I’m hoping to learn something in this thread which I can apply to economics once ems exist, and the above seems to closely correspond with the economic impact of the existence of an em.)
Ahem X-D
When very small shell scripts can be used to replace the researchers who are coming up with ways to keep computer improvements resembling Moore’s Law coming, then that’ll be a bit more relevant. :)
“Computer” used to be a job for humans. As Marc Andreessen pointed out long time ago, software is eating the world—note the present tense, no need to wait for AIs or uploads.
And some jobs are more amenable to replacement than others. (Eg, “Manna”, by Marshall Brain, at http://marshallbrain.com/manna1.htm .) It seems safe to say that being able to create ems would be a significant step in replacing jobs currently held by biological humans; but there are all sorts of details involved which change the economic equations, such as how much RAM an em requires, and when the first em comes online. I’m afraid that your statement doesn’t seem to offer anything that I can use to improve my current estimates on any of these matters.
True. Sorry for the derailing onto a side track :-)
No worries; thread drift happens.
Now, is there any chance I can get you to offer any answers to my questions in my original comment..? ;)
I am still not going to be particularly useful :-/
With respect to uploads/ems my position is hardcore Knightian uncertainty: not only I don’t have any estimates of the timing, I will disbelieve any estimates other people produce as well.
In fact, I don’t know if one can generate an upload by brain scanning at all. I certainly don’t think it’s an inevitability only delayed by the need to develop appropriate tech.