Re: Sentience [Was FAI: Collective Volition]

From: Philip Sutton (Philip.Sutton@green-innovations.asn.au)
Date: Tue Jun 15 2004 - 06:30:32 MDT


Hi John,

> And thatís what I think is so crazy because we come right back to the
> same problem, evolution only sees general problem solving, thatís the
> only thing that enhances survival, so if they (intelligence and
> consciousness) can be separated whatís the point of consciousness?
> Thatís why I said if I though that was true Iíd become a creationist.

I think that the way evolution works, at least in more complex animals,
is that survival-pressure results in a set of capabilities and inbuilt
motivations, and some of these take on a life of their own and become
further developed - to the limit imposed by the need to survive.

Some monkeys (eg. maques) have a rather complex caste system and
the monkeys invest a large amount of their time and brain space to
playing power games or protecting themselves from the impacts of
power games. I think it is arguable that the survival of the *group*
doesn't benefit so much from this activity, but because the power
games are linked to higher level of biological reproduction the genes
that underpin the power game capability and predisposition do well. (ie.
the behaviour favours a subset of genes, not the genes of the whole
clan - arguably the subset of genes are parasitic on the survival
capability of the clan).

In humans we have taken this even further and now in many societies
the power games have become separated from biological reproduction
- effective power-gamers might actually have smaller families (or
none).

So it's possible that consciousness has emerged and developed as a
by-product - first as an enhancement that facilitated adaptibility and
then through reproductively useful mind games and then as a turn-on in
its own right.

If this is so then consciousness is a form of intelligence but not all
intelligence requires consciousness.

Which is all very interesting! What might this mean for FAI and the
collective volition - does an FAI need to be not only intelligent but also
adaptable and creative? If so it might be that it needs or will exhibit
consciousness as a byproduct.

Eliezer wanted the FAI to be NON-conscious so that it didn't create
simulation agents that were alive (ie. clones of real humans) which
would be turned on and off as the FAI explored different issues.

But I'm not sure that this is the right conclusion. If an AGI is NOT
conscious it may not be adaptible and creative at the whole-system
level. In this case the only way for it to come up with the goods on
human volition then is to bury the necessary intelligence below the
whole-system level ie. within the simulation agents. This then creates
'living' simulation artifacts - which is what Eliezer wanted to avoid.

If the creation of 'living' simulation artifacts is to be avoided then the
FAI will need instead to be very creative and adaptible at the whole-
system level - which means that most likely it will be conscious!

Cheers, Philip



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:47 MDT