[ExI] Watson on NOVA

[spike]

On Behalf Of Richard Loosemore
...
>But the only people it has drawn into AI are:
...
>(a) People too poorly informed to understand that Watson represents a
non-achievement ..... therefore extremely low-quality talent, or
...
>(b) People who quite brazenly declare that the field called "AI" is not
really about building intelligent systems, but just futzing around with
mathematics and various trivial algorithms.
...
>Either way, the field loses.
...
>You can discount my opinion if you like, but does it not count for anything
at all that I have been working in this field since I first got 
interested in it in 1980?   This is not armchair theorizing here:  I am just
doing my best to summarize a lot of experience...Richard Loosemore



Richard, your viewpoint as one who has been in the field for a long time is
most valuable.  You and I are actually looking at two very different goals
here, as was pointed out in a previous discussion.  You are shooting for
true AI, but I am not, or at least not immediately.  Reasoning: true AI
leads directly to recursive self-improvement, which leads directly to the
singularity, which presents all kinds of risks (and promise (and risks))
because we don't know how to control it, or even if it is controllable.

On the other hand, Watson isn't going to spontaneously take off and do
whatever a real AI wants to do, any more than a chess algorithm will do
that.  Watson will, however, contribute to our wellbeing here and now, along
with the chess algorithms, and the servant-bot algorithms, the sex-bots, and
all the other non AI applications I can imagine will come along and make our
lives more fun and interesting.

I do not regret all the AI talent that has been siphoned into application
development, for I am in no desperate hurry to create AI.  With our current
level of insight and lack thereof into friendly AI, it looks to me like the
risks may outweigh the benefits, at least to the younger people among us.
Five years ago before my son was born, I would have argued the benefits
outweigh the risks.  Now, I wouldn't say that, or rather I can't say it with
any confidence.  

Recall that nuclear fission was discovered a least a decade before the
engineers developed a practical way to safely control it.  AI is analogous
to nuclear fission, and now is 1937.

You and I do not necessarily disagree, we just have different goals.

spike