It seems I cannot post new threads in the AGI list. I'll post it here: Affordable Supercomputer Power for your MAC MINI or MINI PC through Thunderbolt Technology https://www.facebook.com/notes/juan-carlos-kuri-pinto/affordable-supercomputer-power-for-your-mac-mini-or-mini-pc-through-thunderbolt-/10151447105017712
On Mon, May 6, 2013 at 10:09 AM, Matt Mahoney <[email protected]>wrote: > On Sun, May 5, 2013 at 8:53 PM, Leonardo Stern <[email protected]> > wrote: > > On Fri, May 3, 2013 at 9:28 PM, Matt Mahoney <[email protected]> > wrote: > >> On Fri, May 3, 2013 at 6:09 PM, Leonardo Stern < > [email protected]> wrote: > >> > I agree. But ...the whole point of an AGI singularity is that you > only need to reach humal-level inteligence AGI so AGI can improve itself. > >> > Can you agree that building a human-level AGI will cost around 100 > billion ? > >> > >> No, you need human civilization level AGI. It takes a civilization of > >> billions of people to produce computers and electricity and language > >> and an economy. > > > > > > Ok, let's try to slow down this debate avoiding so I can better > understand your arguments. > > My original email I made 4 assumptions:: > > > > 1 - Human-level AGI can trigger singularity. > > 2 - Human-level AGI costs US$ 100.000.000.000 to develop > > 3 - Human-level AGI costs US$ 1.000.000 / month to operate > > 4 - An human-level AGI can replace 8 workers. > > > > You have disagreed with #1, #2 is agree with your own estimations. > > How about #3 and #4 ? Do you agree with these figures ? > > 4. By definition, human-level AGI would replace 1 worker. But what we > will actually do is build lots of narrow AI that each replace a little > bit of work done by lots of people. Collectively these applications > will replace all human labor. > > The idea of robots with human-like intelligence is a dream by people > who are somehow romantically attracted to the idea and ignore the > reality of economy of scale. People within organizations specialize > because it is more efficient. There is no need for their replacements > to be like humans. One "human equivalent" would mean that the system > saves as much money as one person's wages. > > 3. Assume 1 petaflop and 100 TB memory to do the work of one human > with respect to hard problems like vision, language, art, and > robotics. Figure the initial investment in computing power amortized > over its useful lifetime. By Moore's Law, the hardware loses half its > value over 2 years, which makes its useful lifetime 4 years. If the > initial cost is $10 million, then figure $200K per month. In 2 years > it will be $100K per month. There is also the cost of electricity, > currently $100 per hour ($72K per month). This will also drop with > Moore's Law, but rise as energy becomes more scarce. > > -- > -- Matt Mahoney, [email protected] > > > ------------------------------------------- > AGI > Archives: https://www.listbox.com/member/archive/303/=now > RSS Feed: https://www.listbox.com/member/archive/rss/303/23601136-98835e3f > Modify Your Subscription: > https://www.listbox.com/member/?&; > Powered by Listbox: 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
