[ExI] AI thoughts

efc at swisscows.email efc at swisscows.email
Thu Nov 23 10:38:56 UTC 2023

Hello Jason,

On Wed, 22 Nov 2023, Jason Resch via extropy-chat wrote:

>       > then humans would no longer be in control, even if individual LLMs are no smarter than human engineers).
>       I don't think humans have been in control for a long time, certainly
>       not individual humans, and I just don't believe in "elders" or any
>       other group that is exerting control.  Every single human is like a
>       wood chip on a river.
> I appreciate this analogy very much. I have sometimes thought similar things as well, that the greatest and scariest conspiracy
> theory of all is that no one is in control.

This reminds me of daoist poetry where we are floating on the river.
Best thing we can do is to adapt to the current, instead of tring to
swim against it all our lives.

In terms of conspiracies, I am a firm believer in there being no global
conspiracies. Just like you, I believ in large scale moves and trends,
and I also believe that groups of people try to take advantage of them.
They do not control the trends (as individuals) but try and divert small
currents here and there.

Another reason I do not beleive in global, unified conspiracies is that
people talk. It would be, in my opinion, impossible, to keep such things

> But the river analogy adds another dimension which I think is more correct. We are subject to overarching trends and laws of
> evolution,  technological development, economics, etc. and we individual humans are like cells in this greater organism, all in the
> end replaceable cogs whose presence or absence might make a small difference as to when some inevitable discovery might happen, but
> will not prevent it entirely.

Enter "psycho-history"! ;) I agree. There are fundamental laws that
govern how we work, and this of course influences us as a species. I
always thought about if some kind of law of "unification" or
"centralization" can be verbalized or formalized? It seems, through
history, that we have an innate tendency to try at unify our knowledge,
and that our societies keep getting more and more centralized compared
with individual families or groups on the savannah hundreds of thousands
of year ago.

Then you also have the mystical psycho-analysts who argue that until we
consciously realize and take control over our unconscious drives and
desires, we'll keep making the same mistakes as we always do.

>       > I think depth of reasoning may be one area where the best humans are currently dominant, but a small tweak, such as
>       giving LLMs a working memory and recursion, as well as tightening up their ability to make multiple deductive logical
>       steps/leaps, could quickly change this.
>       Can you make a case that it would be worse than the current situation?
> I don't believe it will, but if tasked to make the case, I would say the greatest present danger is that it amplifies the agency of
> any user. So that ill-willed people might become more destructively capable than they otherwise would (e.g. the common example of a
> lone terrorist leveraging the AI's expertise in biotechnology to make a new pathogen) but the Internet has already done this to a
> lesser extent. I think agency amplification applies to everyone, and since there are more good-intentioned people than
> ill-intentioned ones, any generally-available amplification technology tends to be a net positive for humanity.

I think it is very interesting to think about what current LLMs and
video generation capabilities are doing to drive down the cost of
producing fake news. It will cost next to nothing and flood everything.

Open AI and "containing AI" is just a mirage and every single powerful
nation are likely trying their best to come up with the best AI in order
to out compete the rivals.

I wonder if we'll have some kind of cold war type situation where the
most powerful nations have their own AI:s and the rest of the world then
has to align themselves with them taking one side or the other?

Best regards, 

> Jason 

More information about the extropy-chat mailing list