He said, roughly, that there is a 98% chance of something called "Artificial General Intelligence" being developed, roughly a 95% chance it wipes out humanity and roughly a 1-2% chance it falls into the hands of some narrow elite.
-
Show this thread
-
Replying to @glenweyl
You literally did not understand the numbers I used. I at no point said it had a 95% chance of wiping out humanity. I gave no number whatsoever there. What I said was that the chance of bad actors gaining control of AGI was 1.5 orders of magnitude lower than the risk of... 1/
2 replies 0 retweets 14 likes -
Replying to @ESYudkowsky @glenweyl
...everyone dying because nobody was in control of it. Your grasp of the rest of these ideas is at a similar level, you do not have any idea of what we believe, and I don't think that asking if you could manage to repeat just one sentence back literally was a bad response... 2/
3 replies 0 retweets 12 likes -
Replying to @ESYudkowsky @glenweyl
...and I don't think you *can* understand the actual content of our words unless you can shift to a frame of mind where you can separate value-free empirical questions about AI from their political implications. But it's clear that I am not the right person to... 3/
3 replies 0 retweets 7 likes -
Replying to @ESYudkowsky
There is nothing value free in anything you said any more than there are in “empirical facts” about racial differences in intelligence. I also find your epistemology deeply problematic and confused. I don’t think this means you are incapable of parsing English.
4 replies 0 retweets 7 likes -
Replying to @glenweyl
"There is nothing value-free in what you said" makes me think I literally do not understand how you are using words. I would not be offended if you asked me to repeat something back to you.
3 replies 0 retweets 11 likes -
Replying to @ESYudkowsky @glenweyl
Empirical questions are not wholly value-free because one is free to make choices in how one frames them, and those choices depend on one's values. Probabilities of events are only defined relative to a sample space, for instance. Seehttps://srconstantin.wordpress.com/2015/04/30/choice-of-ontology/comment-page-1/ …
1 reply 4 retweets 16 likes -
This insight is kind of an interpretive gloss on the No Free Lunch Theorem. You can't make a model that's really totally a "value-free" improvement over another: more accurate, for *all* possible minds, in *all* possible situations.
2 replies 2 retweets 9 likes -
Lots of other ways to frame this, of course; it's a Nietzschean insight, it seems related to some postmodernist insights (though I'm less familiar with those.)
1 reply 1 retweet 2 likes -
Of course in practice you are not talking to an arbitrary conceivable mind. You are talking to a human being, in fact a human who has many things in common with yourself, and you can justifiably say "C'mon, we share all the relevant values implicit in the statement I'm making."
1 reply 0 retweets 6 likes
So the interesting question is not "is this or isn't this a value-free discussion" but "ok, so *which* premises implicit in this framing do you not share?" "ok, there are some broad assumptions you might call political that I'm starting with; I'm ok with those; now what?"
-
-
To go back to
@ESYudkowsky's specific comment, for instance, the question "what's more likely, that bad actors take control of an AGI or that it destroys humanity on its own", each of those possible "events" depends on assumptions. We know@glenweyl doubts that AGI is a thing;2 replies 0 retweets 1 like -
he may also doubt that "bad actors" are a thing. (Bad relative to *whom*? Can we assume that the two of you can point at the same group of people and call them "bad actors"? People dispute who counts as "a terrorist" depending on their allegiances.)
1 reply 0 retweets 2 likes - 10 more replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.