And, quite franky, I'm tired of this subject. It's dumb and boring, everyone is warning and fearmongering, and nobody is presenting any facts at all.
And, quite franky, I'm tired of this subject. It's dumb and boring, everyone is warning and fearmongering, and nobody is presenting any facts at all.
>We thus designed a brief questionnaire and distributed it to four groups of experts in 2012/2013. The median estimate of respondents was for a one in two chance that high-level machine intelligence will be developed around 2040-2050, rising to a nine in ten chance by 2075. Experts expect that systems will move on to superintelligence in less than 30 years thereafter. They estimate the chance is about one in three that this development turns out to be ‘bad’ or ‘extremely bad’ for humanity.
http://www.nickbostrom.com/papers/survey.pdf
The people warning about the future of are pretty familiar with AI. Your accusations that they are all idiots who have no understanding of AI is way off the mark.
A number of notable people also signed the future of life institute open letter warning about AI: http://futureoflife.org/misc/open_letter
Deleted comment
As for the warning letter: Asimov and other SF authors have been writing such letters for the longest time, there is nothing new there that hasn't been covered many times over.
In fact, one of the most distressing things I hear from leading AI researchers who've made "AI is nothing to worry about any time soon!" comments is that they do not address any of the specific concerns raised by AI safety advocates. Instead, the most you hear from them is "we're really far away from that!" and "we don't know and I'm sure we'll figure it out when we're near that point."
These comments, from Andrew Ng and the like (insanely brilliant people!) show that they really haven't read Bostrom, etc. Or if they have, they didn't explain much (or get quoted) during interviews. It would make me feel more comfortable with their dismissal if they demonstrated a clear understanding of the arguments being made.
It's like saying, circa 1933: "Thinking about the implications of atomic weapons is a fool's errand, because such things are purely hypothetical at this point."
>According to a 2013 survey of the most cited authors in artificial intelligence, experts expect AI to be able to “carry out most human professions at least as well as a typical human” with a 10% probability by the (median) year 2024, with 50% probability by 2050, and with 90% probability by 2070, assuming uninterrupted scientific progress. Bostrom is less confident than this that AGI will arrive so soon:
>>My own view is that the median numbers reported in the expert survey do not have enough probability mass on later arrival dates. A 10% probability of HLMI [human-level machine intelligence] not having been developed by 2075 or even 2100 (after conditionalizing on “human scientific activity continuing without major negative disruption”) seems too low.
This does not cause me to take them seriously. Unless there is something secret in a government black project, there has been almost no progress on AGI... at all... ever.
We have narrow domain specific algorithms that can ape intelligence in limited domains, but only if they are front loaded by human intelligent designers with a priori knowledge about the meta structure of those domains.
There is one class of algorithms that shows some general learning behavior: evolutionary algorithms. Ironically these are the least favored algorithms by CS AI people. I've heard those who work on them made fun of. It's because while GP/EC shows general ability it does so at such prodigious cost in compute cycles that it takes supercomputing resources to get it to do anything interesting. This makes evolutionary algorithms uncompetitive with fast but narrow search and optimization algorithms designed to solve specific problems.