Yes, well I guess I am. I've already gambled pretty much my whole life on it. Now I'm kinda stuck in that it's one of the few things I have to work on.
Anyways, Matt was saying about the DNA, which is cool I like DNA. However it's also true that we don't need petaflops to do what humans do, perhaps to simulate the brain, though theoretically we could achieve human intelligence with the 100watts or what not. Of course as you may know, my project aims to be quite efficient and portable, being an outgrowth of assembly language in human speakable vocabulary. Anyways in terms of data acquisition, saying it has a cost beyond coding the software for it is somewhat of a joke right? Like seriously, we are living in a "data-deluge" where we are flooded with free and easy to acquire data. Sensors in cell phones, tutorials and how-to's on the internet. Just about anything you can imagine has a tutorial of how to do it on the internet. The main issue is making use of that data in ways a computer can understand. Machine learning is a very efficient way of learning from large amounts of data. Neuro-nets are really a form of statistic, and can be implemented much better. In fact there are many forms of statistic that can be used for machine learning, which computers are much better at performing than humans. Human brains I find are really over-rated, as they have to work with biological substrate, similar to how birds have to as well, so they can't be as efficient or specific as machines. Humans have a lot of background processes, pattern matching and all, which is quite similar to supervised and unsupervised learning, though we may call it "intuition", it is a similar approximation. Using a neuro-net to perform statistics, is much more cumbersome than a classical computer, so we may actually achieve higher rates of efficiency with machines than people. It's similar to how yelling isn't as effective at communicating as writing a letter, or posting on a forum. In conclusion, there are much more efficient ways of achieving cognition, than the simulation of biological processes. On Thu, Apr 4, 2013 at 1:11 AM, Russell Wallace <[email protected]>wrote: > On Thu, Apr 4, 2013 at 3:57 AM, Ben Goertzel <[email protected]> wrote: > >> And we will probably see this happen within the next, say, 3-10 years.... >> But it may happen (shortly) *after* some more modestly funded AGI project >> (OpenCog or something else) demonstrates sufficiently exciting progress to >> really drive home to the "powers that be" that this is an area whose time >> has come... >> > > Yes. For example, nobody funded Marie Curie or Robert Goddard with a > billion dollars. Nobody funded Douglas Lenat until after he had produced AM > and Eurisko off his own bat. > > At the same time, funding wrote off Lenat's life. He was the one man I > call better than I am at what I do, but even after it became clear Cyc > wouldn't work, once he had fifty people under him, walking into work one > day and telling them to pack their bags and go home was something mortal > man would never do, never could. > > If you want funding, you first have to produce results capable of > impressing people who control the money. But be careful that you're willing > to gamble the rest of your life on the thing you get funded for. > *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> > <https://www.listbox.com/member/archive/rss/303/5037279-a88c7a6d> | > 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
