00:00:00
In October, over 850 experts, including
00:00:02
yourself and other leaders like Richard
00:00:04
Branson and Jeffrey Hinton, signed a
00:00:06
statement to ban AI super intelligence
00:00:08
as you guys raised concerns of potential
00:00:10
human extinction.
00:00:11
Because unless we figure out how do we
00:00:14
guarantee that the AI systems are safe,
00:00:17
we're toast.
00:00:18
And you talk about this gorilla problem
00:00:20
as a way to understand AI in the context
00:00:22
of humans.
00:00:22
Yeah. So, a few million years ago, the
00:00:24
human line branched off from the gorilla
00:00:26
line in evolution. And now the gorillas
00:00:28
have no say in whether they continue to
00:00:29
exist because we are much smarter than
00:00:31
they are.
00:00:31
So intelligence is actually the single
00:00:33
most important factor to control planet
00:00:34
earth.
00:00:35
Yep.
00:00:35
But we're in the process of making
00:00:36
something more intelligent than us.
00:00:38
Exactly.
00:00:39
Why don't people stop then?
00:00:41
Well, one of the reasons is something
00:00:42
called the Midas touch. So King Midas is
00:00:44
this legendary king who asked the gods,
00:00:46
can everything I touch turn to gold? And
00:00:48
we think of the Midas touch as being a
00:00:49
good thing, but he goes to drink some
00:00:51
water and the water has turned to gold.
00:00:53
And he goes to comfort his daughter, his
00:00:54
daughter turns to gold. So he dies in
00:00:56
misery and starvation. So this applies
00:00:58
to our current situation in two ways.
00:01:01
One is that greed is driving these
00:01:02
companies to pursue technology with the
00:01:04
probabilities of extinction being worse
00:01:06
than playing Russian roulette. And
00:01:08
that's even according to the people
00:01:09
developing the technology without our
00:01:11
permission. And people are just fooling
00:01:13
themselves if they think it's naturally
00:01:15
going to be controllable.
00:01:16
So if you had a button in front of you
00:01:19
which would stop all progress in
00:01:20
artificial intelligence, would you press
00:01:22
it?
00:01:24
Not yet. I think there's still a decent
00:01:26
chance they guarantee safety, which I
00:01:28
can explain.
