Hi all,

On 5/21/07, Bob Mottram <[EMAIL PROTECTED]> wrote:

The SIAI videos which are up on google so far look ok.  I didn't know
that they were actually trying to *build* an AI, as opposed to just
raising the generally relevant issues.


Just to be clear: At the moment there is no AGI project in the
engineering or detailed-design phase being carried out under the aegis
of the SIAI.

This may be a good opportunity to clarify the current state and
future intentions for R&D at SIAI.

What is going on at SIAI (a small organization) right now is pretty
much two things:

1) raising and exploring the generally relevant issues, as Bob says ... in
particular issues related to ethics, AGI and Singularity

2) fundamental theoretical work on AGI, aimed at one day being able
to start a concrete AGI engineering project founded on sound
mathematical and ethical principles

Both of these are, of course, very important things.

If the current fundraising drive succeeds, a third, closely related
strand of work will start:

3) developing open software tools of use to AGI engineers generally;
and potentially (but not necessarily) of use to SIAI's future in-house
AGI engineering projects

A few months ago SIAI asked me to adopt the (very part time,
so far) role of Director of Research -- which basically meant to
formulate a Research Program for SIAI spanning the 3 areas
above, complementing and going beyond their current and
excellent in-house research program conducted by Eliezer
Yudkowsky and his colleagues.

(As an aside, Novamente's President, Bruce Klein, has also
been helping out SIAI on the "outreach" side, in particular
with organization of the 2007 Singularity Summit which I think
is shaping up into a pretty nice event, see

http://www.singinst.org/events

to sign up for notification when more details are available.)

To see what the SIAI Research Program is about, see

http://www.singinst.org/research/summary

(and explore the links on the left side of the page).

Once sufficient funding is achieved, I will then be charged
with staffing and coordinating the Research Program.

Now, about SIAI and AGI engineering..

There is certainly an option that, in future, the work of Eliezer
Yudkowsky or some other not-yet-hired SIAI researcher will lead
to an in-house SIAI AGI engineering project.  I know that Eliezer
is working toward that goal, though according to a deep and difficult
path involving great ethical care and mathematical rigor.

(His plan, as I understand it, is not to start working on AGI engineering
until he has at least reasonably sound mathematical reasons to
believe his AGI will be beneficial to humans once created and taught.  He
would not be content with having strong intuitive reasons to
believe in the future beneficialness of his AGI; nor with a
plan of experimenting with a pre-human AGI to understand its
properties so as to better come to an understanding of its likely
degree of beneficialness.  If I have mis-stated his perspective I am
sure he will correct me ;-)

There is also an option that, in future, SIAI will wind up working very
closely with university or industry research groups who are doing their
own AGI engineering (such as, for example, Novamente LLC, A2I2,
Stanford AI Lab or Self-Aware Systems, to name just a few groups who
have informal relationships with SIAI already).

And, there is not necessarily a contradiction between SIAI eventually
doing its own AGI engineering, and SIAI partnering with other groups
doing their own AGI engineering.  As an example, if in future SIAI pursued
its
own in-house AGI approach based on Eliezer's approach, but also
partnered with Novamente LLC ... this would not pose any obvious problems
since the two approaches are quite different in nature and orientation.

My own company Novamente LLC does not at this point have any
formal relationship with SIAI, but the possibility of crafting such a
relationship has been discussed.  I think such a relationship makes
sense in principle, since at this stage Novamente LLC does not have
resources to focus as much attention on ethical issues as SIAI does.
Hypothetically, one could see SIAI serving as a sort of "virtual
ethics department" for AGI projects whose leaders
recognize the central importance of AGI ethics and would thus value
the chance to work with a group (SIAI) that has devoted so much effort
just to thinking about AGI ethics.

The touchy thing with this kind of arrangement, of course, is the IP
issues.  But for AGI projects that are academic or relatively open
in their approach to IP, it would seem potentially workable.

Anyway, I hope this email has helped to clarify the current state
and future intentions of SIAI AGI R&D.  Questions are welcome!

Thanks
Ben G

-----
This list is sponsored by AGIRI: http://www.agiri.org/email
To unsubscribe or change your options, please go to:
http://v2.listbox.com/member/?member_id=231415&user_secret=fabd7936

Reply via email to