It's weak AGI, Steve. You have to start somewhere, and you have to start simple. A robot that can navigate any terrain of a given kind is a general machine - something that doesn't exist. An extraordinary breakthrough.
Now here's a simple bet for you - you can't give me a single practical example of what the AGI you are talking about will do. You're talking about this-and-that maths, but you haven't got a clue about any practical problems - demonstrably AGI problems - that your maths would apply to. . From: Steve Richfield Sent: Thursday, July 12, 2012 12:19 AM To: AGI Subject: Re: [agi] Analog Computation Mike, On Wed, Jul 11, 2012 at 3:31 PM, Mike Tintner <[email protected]> wrote: Steve: Until that happens, AGI is a complete non-starter. Steve, This is such cobblers It really hurts to see you guys get involved in extraordinarily intricate discussions about the hypothetical machinery an AGI needs, when you haven't got a clue as to what an AGI does. True. We are all still very much in the dark regarding how a self-organizing process could ever create an intelligent system - which was the very point that I was making. If you have a robot that can do what the simplest organisms do - and that is navigate a strip of rocky ground - *any* strip of such ground within reason - then you have an AGI /sub-AGI. And you'll also have a commercially useful robot. No, you just have "weak" AI, or maybe very weak AGI. When it can figure out some good reason for doing that, it might be interesting. Now that'll certainly be hard, but organisms can do it with just a few hundred or thousand neurons, and no differential equations, maths, or logic - difficult to believe, I sure don't believe this. I believe that neurons are doing some pretty heavy math, and until we understand that math (regardless of whether we ever understand how neurons do it), we don't stand a ghost of a chance. Look at something MUCH simpler - flight. Crude simulation (in wind tunnels) was the ONLY method used from the Wright Brothers on into the Jet era. Now, they are designed in "digital wind tunnels" that STILL simulate. Even something as simple as the shapes of aircraft components remains beyond direct computation short of twiddling a simulation until it seems to work well. The functional characteristics of the components of an intelligent system are probably going to be MUCH more intricate than any aircraft wing, yet SO many people think that they can ignore math, simulation, understanding, an just twiddle their way to success. I have heard NO argument that gives me any reason to expect this to EVER succeed. Not in a year, and not in a century. I know, but it's true. Which is why you will fail. So it won't be *that* hard. This siren song is what is motivating most people who are now looking at AGI. They simply can't see the forest for the trees. Start with the problems an AGI has to solve. This is robotics, hardly even weak AI, let alone even "weak" AGI. What would you need for a simple robotic rover that can navigate any rocky terrain ? (a problem BTW which roboticists are actually addressing right now). I don't see that goal as being anywhere between where we now are, and having true machine intelligence. Please go and work on your simple robotic rover, and leave AGI to others. Steve AGI | Archives | Modify Your Subscription ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-c97d2393 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-2484a968 Powered by Listbox: http://www.listbox.com
