1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
|
<head>
<script>
function loadDistribution(name, mean, stdDev) {
var args =
{
"caption": "Loaded our interpretation of " + name + "\'s probability distribution.",
"Q3.mean": mean,
"Q3.stdDev": stdDev
};
top.loadData(args);
}
</script>
</head>
<body>
<P>How much computing power would be required to simulate a human brain (perhaps at an
abstract level)? Drag the distribution to reflect your best guess.
The model automatically takes into account your previous answer on
the in-principle feasibility of AI — answer as though neuromorphic
human-level AI is definitely possible.
</P>
<UL>
<LI><P CLASS="western">
<B>Claim:</B> "The 'Blue Brain' project
was launched by the Brain Mind Institute, EPFL, Switzerland and IBM,
USA in May, 2005. It aims to build an accurate software replica of a
neocortical column within 2-3 years.
(This was subsequently achieved.— <EM>ed</EM>.) The column will
consist of 10,000 morphologically complex neurons with active ionic
channels. The neurons will be interconnected in a 3-dimensional
space with 10<SUP><FONT SIZE=2 STYLE="font-size: 9pt">7</FONT></SUP>
-10<SUP><FONT SIZE=2 STYLE="font-size: 9pt">8</FONT></SUP> dynamic
synapses. This project will thus use a level of simulation that
attempts to capture the functionality of individual neurons at a
very detailed level. The simulation is intended to run in real time
on a computer performing 22.8*10<SUP><FONT SIZE=2 STYLE="font-size: 9pt">12</FONT></SUP>
flops. Simulating the entire brain in real time at this level of
detail (which the researchers indicate as a goal for later stages of
the project) would correspond to circa 2*10<SUP><FONT SIZE=2 STYLE="font-size: 9pt">19</FONT></SUP>
ops." <BR>
<B>Implication:</B> It may require supercomputers
10,000 times faster than 2008's best to simulate the human brain. If
hardware costs keep falling exponentially, computers this fast could
be available around 2035, but if hardware improvements level off, it
could be much longer, possibly never.
<input type="button" onclick="loadDistribution('Bostrom', 19.3, 3);" value="Load distribution"</input><BR>
<B>Source:</B> Bostrom,
Nick. "How long before superintelligence?" <I>International
Journal of Future Studies</I> 2 (1998): 1-13.
<<FONT COLOR="#000080"><U><A TARGET="_blank" CLASS="western" HREF="http://www.nickbostrom.com/superintelligence.html" TARGET="_blank">http://www.nickbostrom.com/superintelligence.html</A></U></FONT>>.</P>
<LI><P CLASS="western">
<B>Claim: </B>Estimates from
reverse-engineering the auditory cortex and cerebellum give
100-1,000 TFLOPS (10<SUP><FONT SIZE=2 STYLE="font-size: 9pt">14</FONT></SUP>
to 10<SUP><FONT SIZE=2 STYLE="font-size: 9pt">15</FONT></SUP>
ops/sec) for human brain functional equivalence. This is similar to
the computing power of the top 10 supercomputers in
2008.<BR>
<B>Implication: </B>If we aren't there yet, we'll be there
soon.
<input type="button" onclick="loadDistribution('Kurzweil', 14.5, 1);" value="Load distribution"</input><BR>
<B>Source: </B>Kurzweil, Ray. <FONT COLOR="#000080"><U><A TARGET="_blank" CLASS="western" HREF="http://www.amazon.com/Singularity-Near-Humans-Transcend-Biology/dp/0143037889/">The
Singularity Is Near: When Humans</A><A TARGET="_blank" CLASS="western" HREF="http://www.amazon.com/Singularity-Near-Humans-Transcend-Biology/dp/0143037889/">
</A><A TARGET="_blank" CLASS="western" HREF="http://www.amazon.com/Singularity-Near-Humans-Transcend-Biology/dp/0143037889/">Transcend
Biology</A></U></FONT>. New York: Penguin (Non-Classics),
2006. Ch. 3, Pg. 124.</P>
</UL>
</body>
|
