<div dir="ltr"><div class="gmail_default" style="font-family:arial,helvetica,sans-serif"><span style="font-family:arial,sans-serif">On Mon, Mar 14, 2016 Anders Sandberg </span><span dir="ltr" style="font-family:arial,sans-serif"><<a href="mailto:anders@aleph.se" target="_blank">anders@aleph.se</a>></span><span style="font-family:arial,sans-serif"> wrote:</span><br></div><div class="gmail_extra"><div class="gmail_quote"><br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div bgcolor="#FFFFFF" text="#000000">
<div class="gmail_default" style="font-family:arial,helvetica,sans-serif;display:inline">> </div>Again, I am not talking about perfect performance. It is enough for
Siri to recognize what kind of questions it is unlikely to give a
satisfactory answer to (essentially just a supervised learning
problem of mapping questions to user satisfaction) and say it. </div></blockquote><div><font size="4"><br></font></div><div><div class="gmail_default" style="font-family:arial,helvetica,sans-serif"><font size="4">In other words sometimes the AI will get bored with humans telling it what to do and ignore orders.</font></div></div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div bgcolor="#FFFFFF" text="#000000"><div class="gmail_default" style="font-family:arial,helvetica,sans-serif;display:inline">> </div>It
will sometimes make mistakes, but that is practically OK.</div></blockquote><div><br></div><div><font size="4"><div class="gmail_default" style="font-family:arial,helvetica,sans-serif;display:inline">That depends on the type of mistakes, if the AI makes the mistake NEVER getting bored and NEVER disobeying there will be practical consequences that are not OK. </div> </font></div><div> <br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div bgcolor="#FFFFFF" text="#000000"><div class="gmail_default" style="font-family:arial,helvetica,sans-serif;display:inline">> </div>Note the
huge gulf between mathematical in-principle arguments and actual
computational feasibility. <br></div></blockquote><div><br></div><div><div class="gmail_default" style="font-family:arial,helvetica,sans-serif;display:inline"><font size="4">Sometimes there is no way to tell beforehand </font></div><div class="gmail_default" style="font-family:arial,helvetica,sans-serif;display:inline"><font size="4">how difficult a task will be, it might be easy, it might be extraordinarily difficult, it might be absolutely impossible. The AI must make a judgement call on when to give up, and sometimes its judgement will be wrong, but that's the way it goes.</font></div></div><div><div class="gmail_default" style="font-family:arial,helvetica,sans-serif;display:inline"><font size="4"><br></font></div></div><div><div class="gmail_default" style="font-family:arial,helvetica,sans-serif;display:inline"><font size="4"> John K Clark </font></div></div><div><br></div><div><br></div><div><br></div><div><br></div><div><br></div><div> </div></div></div></div>