Quite happy to discuss AGI. I've been thinking, if we could truly mainstream self-driving cars by resolving the remaining issues hampering its maturity, it would be a significant step forward towards achieving AGI feasibility. Should AGI researchers be pooling their talent to that purpose, and/or similar industry problems?
Rob ________________________________ From: Matt Mahoney via AGI <[email protected]> Sent: Thursday, 02 August 2018 7:34 PM To: AGI Subject: Re: [agi] Reality I disagree with most of this. On Wed, Aug 1, 2018 at 7:31 PM Steve Richfield via AGI <[email protected]> wrote: > My AGI-related interest here springs from my observation that nearly > everything people expect from an AGI: > 1. Is well within human problem solving ability. No. Machines already do many things that humans cannot. They have for centuries. > 2. Is absolutely impossible. No. The set of things where people still outperform machines is steadily shrinking. Everything the brain does is computable in principle. > 3. Is SO costly that it isn't worth doing. No, we are already doing it. The ROI on automating human labor is world GDP divided by market interest rates, or about $1 quadrillion. > 4. Would attract others to destroy the AGI. People write malware. Future technology will allow people to create self replicating nanotechnology and engineer lethal and highly contagious pathogens at low cost. That doesn't mean all technology is bad. > 5. Would be overcome by other differently-motivated AGIs. The internet really has no competition. > 6. Requires information not available to a non-human. Automating human labor requires collecting human knowledge. The cost is in the tens or hundreds of trillions of dollars, but not impossible. > 7. Something that an AGI can accomplish that humans cannot accomplish. So > far, I have seen NOTHING that falls into this category. Sure they can. A computer can easily remember a 20 digit number and repeat it back. And other mental feats. > Basic System Analysis Step #1: Detetmine what is expected of the system. > Perhaps you can help here? I outlined the requirements for AGI in http://mattmahoney.net/costofai.pdf > I was serious when I made my reverse Turing test proposal to start a > competition for people to try emulating an AGI, as I expected the falacy or > lack thereof to fall out of such a competition. Until that happens, I just > can't take AGIs serioudly. You mean like a CAPTCHA? Yeah, these will eventually stop working. -- -- Matt Mahoney, [email protected] ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T56889bf27bc7ed87-M622ea01734f37e1a50225a30 Delivery options: https://agi.topicbox.com/groups/agi/subscription
