<html>
<head>
<meta content="text/html; charset=windows-1252"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
On 2016-01-31 18:25, John Clark wrote:<br>
<blockquote
cite="mid:CAJPayv2rxTAu=hLkVmRrQtWGqxVO+kRKPEHeD2-DzDd0kJOPfw@mail.gmail.com"
type="cite">
<meta http-equiv="Content-Type" content="text/html;
charset=windows-1252">
<div dir="ltr">
<div class="gmail_default"
style="font-family:arial,helvetica,sans-serif"><a
moz-do-not-send="true"
href="http://www.wired.com/2016/01/googles-go-victory-is-just-a-glimpse-of-how-powerful-ai-will-be/?mbid=nl_13116"><a class="moz-txt-link-freetext" href="http://www.wired.com/2016/01/googles-go-victory-is-just-a-glimpse-of-how-powerful-ai-will-be/?mbid=nl_13116">http://www.wired.com/2016/01/googles-go-victory-is-just-a-glimpse-of-how-powerful-ai-will-be/?mbid=nl_13116</a></a><br>
</div>
</div>
</blockquote>
<br>
Around here the big topic has been how much this implies about
future AI progress. Much has been made of that domain experts were
suprised by how fast it went, but given Armstrong & Sotala's
results, I would not be too shocked: experts are bad at prediction.
<br>
<br>
I think Eliezer has a relevant point: he is concerned that "Human
neural intelligence is not that complicated and current algorithms
are touching on keystone, foundational aspects of it." - i.e. we may
have found a general tool in deep learning that reduces the "to do"
list of AGI by at least one line (out of an unknown number). <br>
<br>
More practically I think the Wired article gets things right: this
is a big deal commercially. Solving tricky value functions is worth
money - and if they do generalize to hand-eye coordination, then we
will have a practical robot revolution. But even merely software
good value functions might be a huge deal when applied to logistics
and other kinds of planning. Or advertising. <br>
<br>
<br>
<pre class="moz-signature" cols="72">--
Anders Sandberg
Future of Humanity Institute
Oxford Martin School
Oxford University</pre>
</body>
</html>