Re: Group topic: Kurzweil + Singularity.

On 25 Apr, 00:11, HMSBeagle <jsb...@xxxxxxxxxxxxx> wrote:
AFAIK, GOFAI is the class of techniques involving solely the
manipulation of high-level, domain-meaningful symbols.
It is a term used in retrospect to characterize the early attempts at
solving AI problems.
In my understanding, it makes no philosophical claims.

But GOFAI has made all sorts of ludicrous claims. 1) It has said that
all intelligence is symbol manipulation. 2) It has said that language
is entirely incapsulatable in a list of grammer rules. 3) It assumed
that vision was an easy problem worth ignoring. 4) It assumed that
manipulating objects with robot hands was an easy problem worth
ignoring. 5) It said AI is a software problem only. 6) In the 1950s,
someone wrote a rudimentary proof-solving program and then announced
to the press that the first "Thinking Machine" had been built. How
ironic that sounds now!

Ok, I was using the term GOFAI to refer to te class of techniques
while you also include in the term the philosophical/psycological
theories that lead to these techniques.
Both usages are correct, I presume.

- Fusion power

I thought that research on fusion power has been stalled for 50 years
or so, hasn't it?

Stalled? Do you mean in terms of a breakthrough or..?

I'm not an expert but it seems to me that most of "developments" they
had were due to bigger magnets and bigger lasers, and despite these
development, any time they increase the input power, feasible fusion
seems to become harder than they thought before (plasma unstability
issues and such).
I'm skeptical about the possibility to develop fusion power in the
foreseable future.

Well you are making an economic argument that it's cheaper and more
efficient. Will that actually be the case?

Not necessarly. It mostly depends on the technology used to develop
it.

If super-human intelligence is created by pushing to the extreme
limits a current technology (for instance, using a cluster of 1 milion
of ordinary computers to implement something very similar to an
accurate simulation of an human brain and have it learn for 50 years),
then it will be very expensive and it will have little room for
further development, thus it will be probably unactractve for
buisiness.

If super-human intelligence is created using a novel technolgy
(biological neural/conventional computer hybrid, quantum computer,
molecular computer, <fill the blank>) then it could be possible that,
in the initial stages, the technology development is exponential
(because of growing investments and feedback effects) and therefore
the AI quickly outcompetes human intelligence.

Anyhow, this is futurology.

.