<html>
<head>
<meta content="text/html; charset=windows-1252"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
On 2016-04-07 14:57, William Flynn Wallace wrote:<br>
<blockquote
cite="mid:CAO+xQEbzv0hdoH812UZn6AjAX2iL_DD=_i_TZ-1_4zu=7S8vTQ@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:comic sans
ms,sans-serif;font-size:small;color:#000000">Before I reply to
this thread, please John or someone, answer this;</div>
<div class="gmail_default" style="font-family:comic sans
ms,sans-serif;font-size:small;color:#000000"><br>
</div>
<div class="gmail_default" style="font-family:comic sans
ms,sans-serif;font-size:small;color:#000000">Is an AI always
programmed to take the most likely route to success - the
highest probability, once it has considered alternative
routes? People might have other motivations. (to me, this
gets us directly into the free will problem, which y'all might
not want to repeat) <br>
</div>
</div>
</blockquote>
<br>
In the reinforcement learning systems I know the best the actions
undertaken are likely to be best, but there is a chance for
"exploratory" other actions.<br>
<br>
As The Book puts it:<br>
<a class="moz-txt-link-freetext" href="https://webdocs.cs.ualberta.ca/~sutton/book/2/node3.html">https://webdocs.cs.ualberta.ca/~sutton/book/2/node3.html</a><br>
<a class="moz-txt-link-freetext" href="https://webdocs.cs.ualberta.ca/~sutton/book/2/node4.html">https://webdocs.cs.ualberta.ca/~sutton/book/2/node4.html</a><br>
<br>
Things get more complex when the action has parts. <br>
<br>
<pre class="moz-signature" cols="72">--
Anders Sandberg
Future of Humanity Institute
Oxford Martin School
Oxford University</pre>
</body>
</html>