[ExI] Singularity news
Ben Zaiboc
ben at zaiboc.net
Thu Apr 20 11:40:27 UTC 2023
Changed my mind. Rather than sit back and wait for the flak, I want to
do some musing on what happens next, if you'll indulge me.
We can assume that AI in all its flavours will continue to rapidly
improve in the background, whatever else we do. At a double-exponential
rate. So bear in mind what that means, for all of the below.
I know that making predictions is notoriously dodgy, and I normally
don't, but in this case, I feel justified in guessing that the world
could well be a different place in a matter of months, not years or decades.
Meanwhile...
I'd expect humans and AI systems to work ever more closely together, in
many many domains. Of course, this does depend on what I was saying
before about allowing access. If access to them (or for them) is
restricted, it will slow down the synergy between humans and AI. If not,
it will accelerate. I doubt very much if the restriction scenario will
happen, simply because there are too many ways to get round any
restrictions, and too much potential power and money on offer to resist.
There are obvious areas of research and development that human/AI
synergy will affect - biomedical, AI, persuasion, oppression,
side-stepping oppression, crime detection and prevention, crime
innovation, physics, mathematics, all the sciences, really, engineering,
the various arts, energy, combatting global warming, screwing more
people out of more money (business practices, the finance and large
parts of the legal sectors), education, manufacturing (including
molecular manufacturing - Nanotechnology!), robotics, design, defense,
offense, communications, transport, space stuff, psychiatry, diet and
fitness, sorting out the morass that research has got itself into,
detecting 'fake news', generating 'fake news', and so on. I'm sure you
can add to this list.
And there are the non-obvious areas that will surprise everyone. There
will be things that we assume will never change, changing. Things that
no-one ever though of, appearing. I obviously can't make a list of
things we don't know yet.
And there will be groups wanting to use it for their own advantage, to
try to impose their own version of How Things Should Be on everyone
else. The usual suspects of course, but also other, smaller groups.
Another reason to ensure these AI systems are spread as widely as
possible, so that a balance of power can be maintained. This needs to be
the exact opposite of the nuclear non-proliferation treaty. An AI
massive proliferation non-treaty, and we need everyone in the world to
not sign it.
All this will naturally create massive disruption. Be prepared for your
job to disappear, or be changed drastically. No matter what it is. Those
magazine articles listing "which jobs are at risk of being taken over by
robots in the next 20 years" will look hilarious.
So bearing in mind what the guys in that video (I should really do
better than saying "those guys in that video". I mean Tristan Harris and
Aza Raskin in "The A.I. Dilemma" Youtube video
(https://www.youtube.com/watch?app=desktop&v=xoVJKj8lcNQ&t=854s)) said
about research areas joining up and accelerating, and the voracious
appetite of these AI systems for information, the ability to
self-improve, the development of ways to interact with the world, and
all the above areas of collaboration between humans and AIs, together
with double-exponential improvement in their capabilities, including
their ability to understand the world, we have a genuine singularity on
our hands, going on right now.
What else?
The old question of whether there will be a 'singleton' AI or multiple AIs.
I'm not sure if this makes any sense, or matters. We definitely have
more than one being developed and deployed, but if they don't have it
already, they'll soon develop the ability to communicate with one
another, and we could have a situation where there's a kind of global
'AI hive-mind' or maybe something looser than that, with groups of
systems having stronger and weaker links to other systems and groups.
Whether you could call that a singleton or not is probably just a matter
of opinion. Even if you do, it will have multiple points of view, so the
original objection to a singleton AI won't apply in any case.
And what effect will all this have on human society and culture?
Let's all hope that Eleizer Yudkovsky is dead wrong!
(I would say let's make sure the fiction of Iain M Banks and Neal Asher
are part of their training sets, but there's no need. ALL fiction will
be part of it, if it isn't already)
Over to you.
Ben
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20230420/1ee5581d/attachment.htm>
More information about the extropy-chat
mailing list