{"id":2876,"date":"2017-05-29T18:08:08","date_gmt":"2017-05-29T22:08:08","guid":{"rendered":"http:\/\/blogs.ams.org\/blogonmathblogs\/?p=2876"},"modified":"2017-05-29T18:08:08","modified_gmt":"2017-05-29T22:08:08","slug":"best-of-the-bots","status":"publish","type":"post","link":"https:\/\/blogs.ams.org\/blogonmathblogs\/2017\/05\/29\/best-of-the-bots\/","title":{"rendered":"Best of the Bots"},"content":{"rendered":"<p class=\"p1\"><span class=\"s1\">A slew of paint colors named by a neural network, including such gems as \u201cturdly\u201d and \u201crose hork,\u201d made it big last week, with mentions in <a href=\"https:\/\/arstechnica.com\/information-technology\/2017\/05\/an-ai-invented-a-bunch-of-new-paint-colors-that-are-hilariously-wrong\/\"><span class=\"s2\">Ars Technica<\/span><\/a>, <a href=\"http:\/\/www.avclub.com\/article\/its-time-game-neural-network-generated-paint-color-255618\"><span class=\"s2\">The AV Club<\/span><\/a>, and even <a href=\"https:\/\/www.theatlantic.com\/technology\/archive\/2017\/05\/when-a-robot-names-a-new-color-of-paint\/527421\/\"><span class=\"s2\">The Atlantic<\/span><\/a>. But for the story straight from the source, check out <a href=\"http:\/\/lewisandquark.tumblr.com\"><span class=\"s2\">lewisandquark.tumblr.com<\/span><\/a>, the wacky neural network blog of optics researcher Janelle Shane. Her <a href=\"http:\/\/lewisandquark.tumblr.com\/post\/160776374467\/new-paint-colors-invented-by-neural-network\"><span class=\"s2\">original post<\/span><\/a> about paint colors went viral, but I got even more out of <a href=\"http:\/\/lewisandquark.tumblr.com\/post\/160985569682\/paint-colors-designed-by-neural-network-part-2\"><span class=\"s2\">part 2<\/span><\/a>, in which she writes about how she tweaked the algorithms to get better results. (Though \u201ccopper panty\u201d is a marginal improvement at best.)<\/span><\/p>\n<div id=\"attachment_2877\" style=\"width: 3929px\" class=\"wp-caption aligncenter\"><img data-recalc-dims=\"1\" loading=\"lazy\" decoding=\"async\" aria-describedby=\"caption-attachment-2877\" class=\"size-full wp-image-2877\" src=\"https:\/\/i0.wp.com\/blogs.ams.org\/blogonmathblogs\/files\/2017\/05\/klee.jpg?resize=640%2C549\" alt=\"\" width=\"640\" height=\"549\" srcset=\"https:\/\/i0.wp.com\/blogs.ams.org\/blogonmathblogs\/files\/2017\/05\/klee.jpg?w=3919&amp;ssl=1 3919w, https:\/\/i0.wp.com\/blogs.ams.org\/blogonmathblogs\/files\/2017\/05\/klee.jpg?resize=300%2C257&amp;ssl=1 300w, https:\/\/i0.wp.com\/blogs.ams.org\/blogonmathblogs\/files\/2017\/05\/klee.jpg?resize=768%2C659&amp;ssl=1 768w, https:\/\/i0.wp.com\/blogs.ams.org\/blogonmathblogs\/files\/2017\/05\/klee.jpg?resize=1024%2C878&amp;ssl=1 1024w, https:\/\/i0.wp.com\/blogs.ams.org\/blogonmathblogs\/files\/2017\/05\/klee.jpg?w=1280 1280w, https:\/\/i0.wp.com\/blogs.ams.org\/blogonmathblogs\/files\/2017\/05\/klee.jpg?w=1920 1920w\" sizes=\"auto, (max-width: 640px) 100vw, 640px\" \/><p id=\"caption-attachment-2877\" class=\"wp-caption-text\">&#8220;May Picture&#8221; by Paul Klee, who as far as I know did not buy his paint from a neural network.\u00a0Public domain, via the <a href=\"http:\/\/www.metmuseum.org\/art\/collection\/search\/483161\">Metropolitan Museum of Art<\/a>.<\/p><\/div>\n<p class=\"p1\"><span class=\"s1\">My introduction to the machine-generated text genre was the now-dormant <a href=\"http:\/\/kingjamesprogramming.tumblr.com\/\"><span class=\"s2\">King James Programming tumblr<\/span><\/a>, featuring lines from a Markov chain trained on the King James Bible and some computer programming texts. Its most recent contribution was \u201c37:29 The righteous shall inherit the land, and leave it for an inheritance unto the children of Gad according to the number of steps that is linear in b.\u201d Timeless wisdom, to be sure.\u00a0<\/span><\/p>\n<p class=\"p1\"><span class=\"s1\">Neural networks and other machine learning processes are hot right now. Google\u2019s AlphaGo, which uses neural networks to decide which moves to play, is now consistently beating the best Go player in the world (read more about that at <a href=\"https:\/\/mathlesstraveled.com\/2017\/05\/27\/alphago\/\"><span class=\"s2\">The Math Less Traveled by Brent Yorgey<\/span><\/a>), so we humans have lost the edge in basically the last game we were still comparatively good at. I think it\u2019s only prudent for us to keep an eye on what our robot overlords have in store for us as they take over more and more formerly-human tasks.<\/span><\/p>\n<p class=\"p1\"><span class=\"s1\">If the paint colors,\u00a0<a href=\"http:\/\/lewisandquark.tumblr.com\/tagged\/recipes\"><span class=\"s2\">recipes<\/span><\/a>, <a href=\"http:\/\/lewisandquark.tumblr.com\/post\/159062141927\/irish-tune-names-invented-by-neural-network\"><span class=\"s2\">Irish tune names<\/span><\/a>, and <span class=\"s2\"><a href=\"http:\/\/lewisandquark.tumblr.com\/post\/159340939102\/the-neural-network-generated-pickup-lines-that-are\">pickup lines<\/a>\u00a0at Lewis and Quark aren&#8217;t enough for you, there&#8217;s a lot more algorithmic creativity to choose from. At jamesoff.net, you can click until you find a recipe that actually sounds like food. (Let me know if you try grilled coffee, with its ingredients of milk, coffee, mayonnaise, and lambchops.) High noon GMT has something even better than Irish tune names: entire Irish tunes generated by computers.\u00a0<\/span>I&#8217;ve\u00a0really enjoyed seeing experiments involving word2vec, which embeds words as vectors in 300-dimensional space. I first remember learning about it from Jordan Ellenberg&#8217;s blog, and just the other day a friend pointed me to a word2vec <a href=\"https:\/\/llamasandmystegosaurus.blogspot.fr\/2017\/05\/alpha.html\">reinterpretation of Genesis I using only words that begin with &#8220;a<\/a>.&#8221;<\/span><\/p>\n<p class=\"p1\"><span class=\"s1\">For bite-sized chunks of machine learning,\u00a0check out this\u00a0<a href=\"https:\/\/twitter.com\/evelynjlamb\/lists\/bots\"><span class=\"s2\">list of bots<\/span>\u00a0I follow on Twitter<\/a>. I&#8217;m using &#8220;bot&#8221; as sort of a catch-all term there for a lot of different kinds\u00a0of computer-generated tweets.\u00a0<a href=\"https:\/\/twitter.com\/censusAmericans\"><span class=\"s2\">Census Americans<\/span><\/a> tweets census data of randomly selected Americans. <a href=\"https:\/\/twitter.com\/symmetric_curve\"><span class=\"s2\">Symmetric Curves<\/span><\/a> tweets beautiful randomly-generated curves with radial symmetry. <a href=\"https:\/\/twitter.com\/picdescbot\"><span class=\"s2\">Picdescbot<\/span><\/a> tweets the image descriptions it comes up with for random pictures from Wikimedia Commons (with <a href=\"https:\/\/twitter.com\/picdescbot\/status\/869206737190047750\"><span class=\"s2\">varying<\/span><\/a> <a href=\"https:\/\/twitter.com\/picdescbot\/status\/868346126755344384\"><span class=\"s2\">degrees<\/span><\/a> of <a href=\"https:\/\/twitter.com\/picdescbot\/status\/868708475681034242\"><span class=\"s2\">success<\/span><\/a>). All of the bots I follow inject a dose of randomness and usually some levity into my day. So I\u2019d like to thank the bots for amusing me so much as they work toward world domination or their own line of Sherman Williams paints \u2014 whichever comes first.<\/span><\/p>\n<div style=\"margin-top: 0px; margin-bottom: 0px;\" class=\"sharethis-inline-share-buttons\" ><\/div>","protected":false},"excerpt":{"rendered":"<p>A slew of paint colors named by a neural network, including such gems as \u201cturdly\u201d and \u201crose hork,\u201d made it big last week, with mentions in Ars Technica, The AV Club, and even The Atlantic. But for the story straight &hellip; <a href=\"https:\/\/blogs.ams.org\/blogonmathblogs\/2017\/05\/29\/best-of-the-bots\/\">Continue reading <span class=\"meta-nav\">&rarr;<\/span><\/a><\/p>\n<div style=\"margin-top: 0px; margin-bottom: 0px;\" class=\"sharethis-inline-share-buttons\" data-url=https:\/\/blogs.ams.org\/blogonmathblogs\/2017\/05\/29\/best-of-the-bots\/><\/div>\n","protected":false},"author":61,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[366,58],"tags":[697,228,695,696,693,694],"class_list":["post-2876","post","type-post","status-publish","format-standard","hentry","category-data-science","category-mathematics-and-computing","tag-algorithmic-creativity","tag-algorithms","tag-artificial-intelligence","tag-bots","tag-machine-learning","tag-neural-networks"],"jetpack_featured_media_url":"","jetpack_shortlink":"https:\/\/wp.me\/p3tW3N-Ko","jetpack_sharing_enabled":true,"_links":{"self":[{"href":"https:\/\/blogs.ams.org\/blogonmathblogs\/wp-json\/wp\/v2\/posts\/2876","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/blogs.ams.org\/blogonmathblogs\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/blogs.ams.org\/blogonmathblogs\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/blogs.ams.org\/blogonmathblogs\/wp-json\/wp\/v2\/users\/61"}],"replies":[{"embeddable":true,"href":"https:\/\/blogs.ams.org\/blogonmathblogs\/wp-json\/wp\/v2\/comments?post=2876"}],"version-history":[{"count":1,"href":"https:\/\/blogs.ams.org\/blogonmathblogs\/wp-json\/wp\/v2\/posts\/2876\/revisions"}],"predecessor-version":[{"id":2878,"href":"https:\/\/blogs.ams.org\/blogonmathblogs\/wp-json\/wp\/v2\/posts\/2876\/revisions\/2878"}],"wp:attachment":[{"href":"https:\/\/blogs.ams.org\/blogonmathblogs\/wp-json\/wp\/v2\/media?parent=2876"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/blogs.ams.org\/blogonmathblogs\/wp-json\/wp\/v2\/categories?post=2876"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/blogs.ams.org\/blogonmathblogs\/wp-json\/wp\/v2\/tags?post=2876"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}