Timeline for How powerful of a computer do I need to simulate and emulate a human brain?
Current License: CC BY-SA 3.0
7 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| May 20, 2018 at 1:35 | comment | added | JDM-GBG | Take one of the most simple cognitive tasks that a toddler can perform: looking at a picture or a drawing of a familiar object and identifying what it is. The most advanced AI programming techniques we've developed to date can be tricked into mis-identifications that are not merely "off", but spectacularly wrong. evolvingai.org/fooling theverge.com/2017/11/2/16597276/… | |
| Oct 11, 2016 at 5:43 | comment | added | JamieB | The inability to program it is a point that seems glossed over a lot. If we knew how to do it, we could do it today, with technology available at Best Buy. It would just run really really slow until the hardware caught up. That is, if a brain requires X calculations per second to run "real time" and the hardware can only do X/10000, then the brain works, but at 1/10,000th of the speed it should, like trying to search a huge database on an 8086. It'll work, if you have the code to do it; it'll just be slow. What we have today isn't lack of speed, but lack of knowledge on how to build it at all. | |
| Apr 12, 2015 at 19:02 | comment | added | Saidoro | If anything, the flop count they gave for the human brain is comically high. FLOPs measure Floating point operations, a human brain capable of 2.2 exaflops as they claimed would be able to add up 2.2 quintillion 32 bit numbers every second. Which is clearly false. Brains may do 2.2 quintillion things per second, but those things certainly aren't flops. | |
| Apr 11, 2015 at 5:51 | comment | added | JDługosz | That 8Pflops supercomputer from 2011 is blown away by a new shinny 33.8 Pflops Chinese supercomputer (as of August 2014). 15 cents per kWh is high: I'm spending 6, plus distribution costs bringing it to near 11. Put the supercomputer next to the power plant to cut the ddistributioncosts and transmission losses. Put in Iceland, running from geothermal power and ambient cooling. | |
| Feb 9, 2015 at 0:22 | comment | added | ckersch | Petaflop estimates for the human brain vary significantly based on source. Most estimates are between 2 and 50 petaflops, so my answer could be off by a factor of about 20. | |
| Feb 9, 2015 at 0:11 | comment | added | user3652621 | Are you confident of that source? The petaflop count depicted in the graph for a human brain seems a bit low... | |
| Feb 8, 2015 at 21:04 | history | answered | ckersch | CC BY-SA 3.0 |