How big is an AGI
57 views
Skip to first unread message
Ed Pell
Mar 16, 2017, 3:25:42 PM3/16/17
to opencog
How many person years does it take to implement the outlines of an AGI system? Thanks.
Linas Vepstas
Mar 16, 2017, 9:08:52 PM3/16/17
to opencog
We here in opencog-land need some dozens of man-years to understand and cross the next few hurdles. In other words, I have a good idea of what code needs to be written, and what research needs to be performed, for the next two cycles. I cannot see past that, because what follows depends on the success/failure of the research that has not been performed.
Its very easy to insist that there are only a few hurdles more, but it has seemed like there were only a few hurdles more since the 1970s, The track-and-field record is poor.
--linasIts very easy to insist that there are only a few hurdles more, but it has seemed like there were only a few hurdles more since the 1970s, The track-and-field record is poor.
On Fri, Mar 17, 2017 at 3:25 AM, Ed Pell <edp...@gmail.com> wrote:
How many person years does it take to implement the outlines of an AGI system? Thanks.
--
You received this message because you are subscribed to the Google Groups "opencog" group.
To unsubscribe from this group and stop receiving emails from it, send an email to opencog+unsubscribe@googlegroups.com.
To post to this group, send email to ope...@googlegroups.com.
Visit this group at https://groups.google.com/group/opencog.
To view this discussion on the web visit https://groups.google.com/d/msgid/opencog/4745dbeb-601d-4ed8-8fb5-6011e3c00578%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
Ed Pell
Mar 17, 2017, 11:16:48 AM3/17/17
to opencog, linasv...@gmail.com
Linas, thanks for your thoughtful reply. Yes, the track record is spotty. I also think that even if a whole system is written it will show/require hardware acceleration for the memory. That is a hardware based associative memory. That will allow mulling over big chucks of the memory in a reasonably short time. I expect once an architecture for such a memory has reached a consensus level it will be quick for a start up to produce.
If only there were a large corporation or military research organization willing to pay for a 100 person five year effort.
I see endless folks nibbling in from the edges. That is sensors and services but not core, no understanding, reasoning, learning.
Neural networks are useful when implemented in cmos (not software emulation) to abstract up raw audio and video but again at the edges.
If only there were a large corporation or military research organization willing to pay for a 100 person five year effort.
I see endless folks nibbling in from the edges. That is sensors and services but not core, no understanding, reasoning, learning.
Neural networks are useful when implemented in cmos (not software emulation) to abstract up raw audio and video but again at the edges.
Reply all
Reply to author
Forward
0 new messages