From: Richard Loosemore (firstname.lastname@example.org)
Date: Wed Sep 07 2005 - 08:12:59 MDT
I want to add my voice in support of what Justin and Ben are saying
here, and add another thought.
I was in the middle of my Ph.D. thesis (in AI/Psychology) in 1992 when I
finally realised that the systems I wanted to build *could not* be built
without incredibly sophisticated tools. I stopped work and switched
careers to become a software engineer for a while, with the aim of
finding out just what kind of tools would be needed.
I am now firmly convinced that not only can one or two people not
succeeed in the task of building an AGI, but that a team of N people
where N > 0 cannot do it, unless there is a serious revolution in the
art of software tool building.
The revolution I have in mind (and I can hear the groans starting even
as I write these words) involves the introduction of serious amounts of
"psychology of programming" into the field of tool building. I use
scare quotes around that phrase because I specifically do not mean the
kind of psychology of programming that exists at the moment. Quite
simply, the way that programmers (and all others involved) approach the
task of building large systems is based on ideas that were appropriate
for small single-processor systems, and completely inappropriate for
massively parallel systems. No account I taken of the specific fatcs
about the limitations of human minds, and until someone realises that
the path to large system building runs through a preliminary period of
revolutionary tool building, no progress is going to be made.
I say all this mindful of the fact that there are some radical projects
going on at the moment. I have looked at these, and from what I see,
the teams, though briliant, do not know anything about psychology. I
have no reason to be optimistic about those efforts.
Ben Goertzel wrote:
> Justin Corwin wrote:
>>can't think of any one person who can seriously be credited not only
>>with innovative theories, but a complex final design of a nontrivial
>>part of our world. Things are much too complicated for that.
>>Especially software design, wherein the interactions (even inside a
>>system that could be said to be designed by one person) are often
>>entirely unexpected to the designer.
>>I suppose it just seems very unrealistic to imagine that all the
>>relevant complexity could be solved, designed, communicated to drones
>>or done personally, all by one person in any kind of reasonable
>>time-scale, if it were possible at all. AI in particular is not just
>>an architectural and theoretical problem, but many many
>>implementation, engineering, and "mere" software design issues.
> I agree completely.
> Given the current state of hardware/software tech, it is not feasible
> for one person to put together a seed AI.
> And it's not feasible to envision seed AI as cleanly dividing into an
> "AI layer" and an "implementation layer", with the maverick genius
> doing the former and his drones filling in the latter.
> At the present time, there needs to be a fair bit of feedback between
> these two layers of any wannabe-AGI system, even if they are *reasonably*
> well separated.
> On the other hand, this may not be the case in another 15 years, once
> hardware and software have advanced considerably.
> So what I would say is: If some small team of individuals doesn't create
> AGI in the next 5-20 years, then perhaps some lone maverick will do it
> 20-40 years from now, once technology has developed far enough to support
> lone-maverick activities in this area.
> The reason we don't see lone mavericks building significant, large,
> real-world systems is that there are a lot of people in the world, and
> teams generally emerge to do the job before the underlying support tech
> reaches the point to enable individual-maverick activity.
>>To load examples a bit, I could hold up Google, ostensibly a simple
>>innovative idea by two graduate students. But upon closer inspection,
>>the actual reality of google is the interaction of much more than
>>that. Sergey and Larry did not design the cluster architecture it runs
>>on, nor the Google File System which contains their database, nor the
>>MapReduce implementation that underlies their fast response time.
> Now, in 30 years, implementing Google may be a one-man job, because all
> these support technologies will be commoditized. But there are many
> enterprising humans, so many valuable things get done when they're still
> at the "needs a team to do it" stage, in terms of the availability of
> underlying support technologies.
>>Quite aside from the time aspect, a single person is simply
>>too prone to conceptual mistakes and self induced theory blocks. As
>>optimistic as I am about mental plasticity, I don't' think any one
>>person can really solve all the problems that are theoretically within
>>their intelligence and ability. People get stuck, or lack inclination,
>>or prefer to focus on more interesting problems, or simply run out of
> I think this is a far weaker point.
> While what you say is statistically true, there may always be exceptions
> to this sort of rule....
> I am much more in sympathy with your prior comments about
> a) the sheer mass
> of details involved in a large-scale engineering project making use of
> immature technologies
> b) the difficulty of separating the purely-scientific from engineering-
> oriented aspects of a complex engineering project
> -- Ben
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:52 MDT