
<html><head></head><body><div><span data-mailaddress="protokol2020@gmail.com" data-contactname="Tomaz Kristan" class="clickable"><span title="protokol2020@gmail.com">Tomaz Kristan</span><span class="detail"> <protokol2020@gmail.com></span></span> , 7/9/2014 3:48 PM:</div><blockquote class="mori" style="margin: 0px 0px 0px 0.8ex; border-left-width: 2px; border-left-color: blue; border-left-style: solid; padding-left: 1ex;"><div><br></div><div>10 years ago I contemplated and co-wrote an algorithm evolver, you can download it here:</div><div><br></div><div>http//<a href="http://www.critticall.com/Setup_Critticall137.exe" title="http://www.critticall.com/Setup_Critticall137.exe" target="_blank">www.critticall.com/Setup_Critticall137.exe</a> <br></div><div><br></div><div>It can do quite non-trivial things.</div></blockquote><div><br></div>Looks neat. Have you written up something about it? The demonstrations look good (love those packings) but it would be nice to be able to read how it works.<div><br></div><div>Overall, genetic approaches to optimization are nice. But they mainly work when the fitness landscape is somewhat forgiving. That is why optimizing algorithms themselves is so tricky, and every little step forward is worth investigating deeply.</div><div><br><blockquote class="mori" style="margin: 0px 0px 0px 0.8ex; border-left-width: 2px; border-left-color: blue; border-left-style: solid; padding-left: 1ex;"><div></div></blockquote><br>Anders Sandberg,

Future of Humanity Institute

Philosophy Faculty of Oxford University</div></body></html>