Eliezer's probably not going to be available over the next 4 days due to
the Foresight Gathering, so I'll take a shot at this. Flare actually is
on our list of things to do right after finishing Friendly AI. So we hope
to get some work done on that this summer. However, it will only be the
preliminary work- making the spec, etc. etc. The idea after that is to
hand it off to a full time person to manage, and that person can bring in
other people to work part time on it if they want. If we found a full time
volunteer we definitely would consider that.
More importantly, we do not have any plans to rely on Flare... at this
point it's more of a "nice if you got it, but not absolutely necessary".
Late this year we plan to begin coding work on our AI, and at the current
time it is looking like we will do the initial work in Python. If at some
time later Flare becomes usable we would switch over. Or if Flare never
gets built we can probably struggle along with just using Python and other
Many of the documents at sysopmind are out of date in parts, so beware.
Anything not updated since last fall is most definitely out of date. The
only stuff you can definitely count on being up to date are documents hosted
directly at SingInst.org.
Adrian Tymes wrote:
> Nice idea, but - from reading TPtS (for those who haven't read: it's
> linked off of CaTAI, and basically the "so how do we do it?" part), it
> sounds like the seed somehow requires Flare as Eleizer has dreamed it
> up, to go beyond initial stages. (At least, crediting it with being
> able to "create the potential for the timeline" that leads to an AI,
> and similar comments, is pretty suggestive in that regard IMO.)
> Without the specs for Flare, it is difficult to see how Flare solves
> the problems it is credited with solving, yet the plan assumes that
> Flare has solved them.
-- Brian Atkins Director, Singularity Institute for Artificial Intelligence http://www.singinst.org/
This archive was generated by hypermail 2b30 : Mon May 28 2001 - 09:59:49 MDT