Think Steve has summed this up pretty well at this point. I'm not going to go into the consciousness arguments and computability. I don't have the bandwidth.
-GJS On Sat, Sep 19, 2015 at 2:10 PM, Steve Richfield <[email protected]> wrote: > Ed, > > <- snip -> > There is a good chance we will be able to build computers having many of > these beneficial properties of the brain within 5 to 15 years. > > Y'know, I once believed much the same, but experience has taught me > better. In 1980 I projected that by 2000 that it would be possible to > purchase a computer to simulate a person, neuron by neuron, synapse by > synapse, in real time for ~$5K. > > Then, when this didn't happen, I CAREFULLY analyzed why. Nearly all of my > error was in predicting major advances in architecture that just didn't > happen. To make my point I looked at presently known architectural advances > overlaid on 2000 fab methods, and it comes pretty close. I wrote a paper > about what it would take that I labeled *The Itanium Effect* without > reference to comparing computational abilities to humans. This would > provide ~2 orders of magnitude in performance on PRESENT fabs, but MUCH > more on 2000 fabs. Still, there was a piece missing, so... > > My 1980 calculations included the presumption that neurons would be > computed on an as-needed basis rather than being constantly recomputed, > which is worth ~4 orders of magnitude. Of course the challenge is in > figuring out when re-computation is needed without eating up the benefits. > Then I wrote DrEliza.com and ran into a parallel problem, which I solved in > a recently-issued patent discussed earlier on this forum. The method > involved bottom-level triggering that looks transportable to neuromorphic > methods. I can envision a new op-code that would seem do this. > > So, there seems to be a half dozen orders of magnitude available in KNOWN > approaches - yet NOTHING is happening "on the ground". Why? > > I think (but am not sure) I understand. Many of these issues are SUBTLE, > SO subtle that people on this forum are now writing code that absolutely > can NOT work (in real time) despite having discussed these issues here. The > few people who grok these issues simply disappear from this forum because > they can't see making anything work in the present environment. In a real > sense the AGI Rapture has already happened, and those who grok its likely > construction have already "risen" above this forum, leaving those who just > can't get it, and a few of us "in betweeners" who weren't good enough to > rise, to try to convince the Left Behind of the nature of the situation. > > Meanwhile, manufacturers see NO market for such hardware, especially no > market to support the ~half-billion dollars it would take to bring such a > processor to market. > > This field is DEAD DEAD DEAD until someone with a LOT of money WAKES UP. > No corporation can take this step because they are structurally incapable > to making such giant leaps into marketing terra incognita. > > So, don't expect to see this in another 15 years UNLESS someone wakes up. > It hasn't happened yet, and I don't see it happening in the future. > Something happens to people when they get a LOT of money. They think the > limits to technology they see are there to stay. They lose their sense of > futurism. > > The closest "near miss" seems to be Paul Allen, who is pumping a LOT of > money into scanning brains, but at too low of a resolution to be worth > anything for our purposes. He could pump some of that money into microscope > development, but instead apparently accepts the crap that he can purchase. > > What is needed here is a route past the above described roadblock. The > first step would seem to be to outline the possible routes. Does anyone > here see a possible route past this roadblock? > > Steve > > *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> > <https://www.listbox.com/member/archive/rss/303/27055757-c218d4f9> | > Modify > <https://www.listbox.com/member/?&;> > Your Subscription <http://www.listbox.com> > ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657 Powered by Listbox: http://www.listbox.com
