{"id":35455,"date":"2018-11-13T19:17:54","date_gmt":"2018-11-14T00:17:54","guid":{"rendered":"https:\/\/digital.hbs.edu\/platform-rctom\/submission\/dream-on-an-exploration-of-neural-nets-turned-inside-out\/"},"modified":"2018-11-13T19:17:54","modified_gmt":"2018-11-14T00:17:54","slug":"dream-on-an-exploration-of-neural-networks-turned-inside-out","status":"publish","type":"hck-submission","link":"https:\/\/d3.harvard.edu\/platform-rctom\/submission\/dream-on-an-exploration-of-neural-networks-turned-inside-out\/","title":{"rendered":"Dream On: An Exploration of Neural Networks Turned Inside Out"},"content":{"rendered":"

Human consciousness is likely one of the most significant marvels of nature in the entire universe.\u00a0 With all our technology, scientific advancement, and ever-increasing computational power, we still have so little understanding of how our own minds actually work.\u00a0 In the quest to deepen our understanding, the field of machine learning and artificial intelligence is growing by leaps and bounds.\u00a0 Already, it has enabled things that seemed like magic a few years ago\u2014Alexa can tell you jokes, AI chatbots can answer your questions on Facebook Messenger, and Google\u2019s advanced neural-net machine learning algorithms can recognize pictures of virtually anything.\u00a0 In a fascinating twist, however, it turns out that the inner workings of neural networks are so complicated that no one actually understands how they\u00a0work\u00a0<\/em>[1].\u00a0 It was while pondering this question that Google engineers had a crazy idea\u2014they decided to let the algorithm dream.<\/p>\n

Now known as DeepDream, Google has taken their neural net image recognition software and programmed it to run backward.\u00a0 Artificial neural networks work by being trained on millions of examples and gradually adjusting network parameters until they result in the classification desired [1].\u00a0 For example, a net might be shown millions of pictures of dogs until it is able to recognize dogs with a high degree of accuracy.\u00a0 Due to the complexity of the learning process, however, what is actually happening inside the neural net is poorly understood.\u00a0 To understand this process better, Google engineers decided to turn the operation inside out.<\/p>\n

Initially, engineers fed the neural net an image of a random scene and allowed it to look for an image to recognize.\u00a0 Once the program came back with a fit, they then instructed it to change<\/em> the original image to fit the program\u2019s classification algorithm better.\u00a0 Repeating this cycle many times resulted in psychedelic images \u201cdreamed\u201d by the algorithm\u2014essentially manifestations of what the algorithm understood the classification in question to be [2].<\/p>\n

Figure 1: \u201cDreaming\u201d Process<\/u><\/p>\n

\"\"<\/a><\/p>\n

Source: [2]<\/p>\n

Figure 2:\u00a0 50 iterations of \u201cdreaming\u201d with an algorithm trained on dogs, given an image of jellyfish<\/u><\/p>\n

\"\"<\/a><\/p>\n

Source: [3]<\/p>\n

Taking it a step further, Google then tested their neural net on pure white noise.\u00a0 Given images with absolutely no patterns, this allowed the algorithm to dream much like a human does when looking at clouds.\u00a0 When we see a shape that is similar to something we recognize, we might say a cloud looks like a rabbit or a car.\u00a0 This abstract generalization is formed in the depths of our consciousness [4].\u00a0 By replicating this process with DeepDream, Google is gaining insight into how its algorithm thinks<\/em>.<\/p>\n

Figure 3: Images generated from random noise<\/u><\/p>\n

\"\"<\/a><\/p>\n

Source: [1]<\/p>\n

In some cases, the results are surprising!\u00a0 According to a Google researcher, \u201cThe problem is that the knowledge gets baked into the network, rather than into us [4].\u201d\u00a0 In one particularly interesting case, a program \u201cdreaming\u201d dumbbells produced images of dumbbells with arms attached to them!\u00a0 Since the network had likely rarely seen images of dumbbells without an arm lifting them, it associated the arm with the object itself [1].\u00a0 Examples like this show the power of this analysis\u2014Google has found a way to see into the mind of their creation and understand how a program understands<\/em> reality.\u00a0 In this case, the algorithm failed to separate the weight from the weight-lifter!<\/p>\n

Figure 4: Dumbbells with arms attached<\/u><\/p>\n

\"\"<\/a><\/p>\n

Source: [1]<\/p>\n

\u00a0The implications of this technology are profound.\u00a0 In the short-term Google will be able to leverage this type of processing to improve their existing image recognition software and increase their understanding of how neural nets learn and understand reality.\u00a0 This will benefit consumers as AI products become increasingly accurate and begin to truly understand the world at contextually-indifferent depths [4].\u00a0 The longer-term implications are even more exciting, however.\u00a0 Eventually, this understanding may lead to software that is predictive.\u00a0 In other words, algorithms may learn to fill in missing pieces because they understand<\/em> what is missing [2].\u00a0 This opens up a world of applications, from automatic photoshopping to the mythical \u201czoom and enhance\u201d of sci-fi movies [3].\u00a0 Some even speculate that DeepDream may be able to create testable predictions for the pathogenesis of psychosis by understanding how false perceptions of reality arise in the mind [5].\u00a0 Personally, in the next several years I would love to see Google explore applications in expressive music generation, creative art, and even storytelling.<\/p>\n

DeepDream is a fascinating step in understanding how computer learning algorithms understand reality.\u00a0 Hopefully, this understanding will eventually shed light on the great mystery of our own cognition.\u00a0 However, many questions still remain.\u00a0 Do our minds actually function in a meaningfully similar way to neural nets?\u00a0\u00a0 If machines eventually \u201cthink\u201d and \u201cunderstand,\u201d will they understand in a way similar enough to humans to be even recognizable to our organic minds?\u00a0 How do we separate machine learning insights from garbage when the output is so profound that we ourselves don\u2019t even recognize its brilliance?\u00a0 I don\u2019t know the answers to these questions, but I certainly dream of a marvelous future.<\/p>\n

 <\/p>\n

Word Count: 798<\/u><\/p>\n

 <\/p>\n

Citations:<\/u><\/strong><\/p>\n

    \n
  1. Alexander Mordvintsev, et al., \u201cInceptionism: Going Deeper into Neural Networks,\u201d Google AI Blog, June 17, 2015, https:\/\/ai.googleblog.com\/2015\/06\/inceptionism-going-deeper-into-neural.html<\/a>, accessed November 2018.<\/li>\n
  2. Wilkin, Henry, \u201cPsychosis, Dreams, and Memory in AI,\u201d Special Edition on Artificial Intelligence (blog), August 28, 2017, http:\/\/sitn.hms.harvard.edu\/flash\/2017\/psychosis-dreams-memory-ai\/<\/a>, accessed November 2018.<\/li>\n
  3. Wikipedia, \u201cDeepDream,\u201d https:\/\/en.wikipedia.org\/wiki\/DeepDream<\/a>, accessed November 2018.<\/li>\n
  4. Castelvecchi, D, \u201cCan we open the black box of AI?,\u201d\u00a005 October 2016, Nature News<\/em>,\u00a0538<\/em>(7623), p.20.<\/li>\n
  5. Keshavan, Matcheri S. et al. Deep dreaming, aberrant salience and psychosis: Connecting the dots by artificial neural networks<\/a>. Schizophrenia Research , Volume 188 , 178 \u2013 181.<\/li>\n<\/ol>\n","protected":false},"excerpt":{"rendered":"

    What if computers could dream? In fact, they can. Google’s groundbreaking DeepDream software is turning AI neural networks inside out to understand how computers think.<\/p>\n","protected":false},"author":11573,"featured_media":35456,"comment_status":"open","ping_status":"closed","template":"","categories":[1869,278,1909,2061,346,2461,1015],"class_list":["post-35455","hck-submission","type-hck-submission","status-publish","has-post-thumbnail","hentry","category-ai","category-art","category-artificial-intelligence","category-data-visualization","category-machine-learning","category-machine-vision","category-predictive-technology","hck-taxonomy-organization-harvard-business-school","hck-taxonomy-industry-technology","hck-taxonomy-country-united-states"],"connected_submission_link":"https:\/\/d3.harvard.edu\/platform-rctom\/assignment\/rc-tom-challenge-2018\/","yoast_head":"\nDream On: An Exploration of Neural Networks Turned Inside Out - Technology and Operations Management<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/d3.harvard.edu\/platform-rctom\/submission\/dream-on-an-exploration-of-neural-networks-turned-inside-out\/\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Dream On: An Exploration of Neural Networks Turned Inside Out - Technology and Operations Management\" \/>\n<meta property=\"og:description\" content=\"What if computers could dream? In fact, they can. Google's groundbreaking DeepDream software is turning AI neural networks inside out to understand how computers think.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/d3.harvard.edu\/platform-rctom\/submission\/dream-on-an-exploration-of-neural-networks-turned-inside-out\/\" \/>\n<meta property=\"og:site_name\" content=\"Technology and Operations Management\" \/>\n<meta property=\"og:image\" content=\"https:\/\/d3.harvard.edu\/platform-rctom\/wp-content\/uploads\/sites\/4\/2018\/11\/mike-tyka4.jpg\" \/>\n\t<meta property=\"og:image:width\" content=\"700\" \/>\n\t<meta property=\"og:image:height\" content=\"467\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/jpeg\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data1\" content=\"5 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/d3.harvard.edu\\\/platform-rctom\\\/submission\\\/dream-on-an-exploration-of-neural-networks-turned-inside-out\\\/\",\"url\":\"https:\\\/\\\/d3.harvard.edu\\\/platform-rctom\\\/submission\\\/dream-on-an-exploration-of-neural-networks-turned-inside-out\\\/\",\"name\":\"Dream On: An Exploration of Neural Networks Turned Inside Out - Technology and Operations Management\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/d3.harvard.edu\\\/platform-rctom\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/d3.harvard.edu\\\/platform-rctom\\\/submission\\\/dream-on-an-exploration-of-neural-networks-turned-inside-out\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/d3.harvard.edu\\\/platform-rctom\\\/submission\\\/dream-on-an-exploration-of-neural-networks-turned-inside-out\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/d3.harvard.edu\\\/platform-rctom\\\/wp-content\\\/uploads\\\/sites\\\/4\\\/2018\\\/11\\\/mike-tyka4.jpg\",\"datePublished\":\"2018-11-14T00:17:54+00:00\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/d3.harvard.edu\\\/platform-rctom\\\/submission\\\/dream-on-an-exploration-of-neural-networks-turned-inside-out\\\/#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/d3.harvard.edu\\\/platform-rctom\\\/submission\\\/dream-on-an-exploration-of-neural-networks-turned-inside-out\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/d3.harvard.edu\\\/platform-rctom\\\/submission\\\/dream-on-an-exploration-of-neural-networks-turned-inside-out\\\/#primaryimage\",\"url\":\"https:\\\/\\\/d3.harvard.edu\\\/platform-rctom\\\/wp-content\\\/uploads\\\/sites\\\/4\\\/2018\\\/11\\\/mike-tyka4.jpg\",\"contentUrl\":\"https:\\\/\\\/d3.harvard.edu\\\/platform-rctom\\\/wp-content\\\/uploads\\\/sites\\\/4\\\/2018\\\/11\\\/mike-tyka4.jpg\",\"width\":700,\"height\":467},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/d3.harvard.edu\\\/platform-rctom\\\/submission\\\/dream-on-an-exploration-of-neural-networks-turned-inside-out\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/d3.harvard.edu\\\/platform-rctom\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Submissions\",\"item\":\"https:\\\/\\\/d3.harvard.edu\\\/platform-rctom\\\/submission\\\/\"},{\"@type\":\"ListItem\",\"position\":3,\"name\":\"Dream On: An Exploration of Neural Networks Turned Inside Out\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/d3.harvard.edu\\\/platform-rctom\\\/#website\",\"url\":\"https:\\\/\\\/d3.harvard.edu\\\/platform-rctom\\\/\",\"name\":\"Technology and Operations Management\",\"description\":\"MBA Student Perspectives\",\"potentialAction\":[{\"@type\":\"性视界Action\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/d3.harvard.edu\\\/platform-rctom\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Dream On: An Exploration of Neural Networks Turned Inside Out - Technology and Operations Management","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/d3.harvard.edu\/platform-rctom\/submission\/dream-on-an-exploration-of-neural-networks-turned-inside-out\/","og_locale":"en_US","og_type":"article","og_title":"Dream On: An Exploration of Neural Networks Turned Inside Out - Technology and Operations Management","og_description":"What if computers could dream? In fact, they can. Google's groundbreaking DeepDream software is turning AI neural networks inside out to understand how computers think.","og_url":"https:\/\/d3.harvard.edu\/platform-rctom\/submission\/dream-on-an-exploration-of-neural-networks-turned-inside-out\/","og_site_name":"Technology and Operations Management","og_image":[{"width":700,"height":467,"url":"https:\/\/d3.harvard.edu\/platform-rctom\/wp-content\/uploads\/sites\/4\/2018\/11\/mike-tyka4.jpg","type":"image\/jpeg"}],"twitter_card":"summary_large_image","twitter_misc":{"Est. reading time":"5 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/d3.harvard.edu\/platform-rctom\/submission\/dream-on-an-exploration-of-neural-networks-turned-inside-out\/","url":"https:\/\/d3.harvard.edu\/platform-rctom\/submission\/dream-on-an-exploration-of-neural-networks-turned-inside-out\/","name":"Dream On: An Exploration of Neural Networks Turned Inside Out - Technology and Operations Management","isPartOf":{"@id":"https:\/\/d3.harvard.edu\/platform-rctom\/#website"},"primaryImageOfPage":{"@id":"https:\/\/d3.harvard.edu\/platform-rctom\/submission\/dream-on-an-exploration-of-neural-networks-turned-inside-out\/#primaryimage"},"image":{"@id":"https:\/\/d3.harvard.edu\/platform-rctom\/submission\/dream-on-an-exploration-of-neural-networks-turned-inside-out\/#primaryimage"},"thumbnailUrl":"https:\/\/d3.harvard.edu\/platform-rctom\/wp-content\/uploads\/sites\/4\/2018\/11\/mike-tyka4.jpg","datePublished":"2018-11-14T00:17:54+00:00","breadcrumb":{"@id":"https:\/\/d3.harvard.edu\/platform-rctom\/submission\/dream-on-an-exploration-of-neural-networks-turned-inside-out\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/d3.harvard.edu\/platform-rctom\/submission\/dream-on-an-exploration-of-neural-networks-turned-inside-out\/"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/d3.harvard.edu\/platform-rctom\/submission\/dream-on-an-exploration-of-neural-networks-turned-inside-out\/#primaryimage","url":"https:\/\/d3.harvard.edu\/platform-rctom\/wp-content\/uploads\/sites\/4\/2018\/11\/mike-tyka4.jpg","contentUrl":"https:\/\/d3.harvard.edu\/platform-rctom\/wp-content\/uploads\/sites\/4\/2018\/11\/mike-tyka4.jpg","width":700,"height":467},{"@type":"BreadcrumbList","@id":"https:\/\/d3.harvard.edu\/platform-rctom\/submission\/dream-on-an-exploration-of-neural-networks-turned-inside-out\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/d3.harvard.edu\/platform-rctom\/"},{"@type":"ListItem","position":2,"name":"Submissions","item":"https:\/\/d3.harvard.edu\/platform-rctom\/submission\/"},{"@type":"ListItem","position":3,"name":"Dream On: An Exploration of Neural Networks Turned Inside Out"}]},{"@type":"WebSite","@id":"https:\/\/d3.harvard.edu\/platform-rctom\/#website","url":"https:\/\/d3.harvard.edu\/platform-rctom\/","name":"Technology and Operations Management","description":"MBA Student Perspectives","potentialAction":[{"@type":"性视界Action","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/d3.harvard.edu\/platform-rctom\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"}]}},"_links":{"self":[{"href":"https:\/\/d3.harvard.edu\/platform-rctom\/wp-json\/wp\/v2\/hck-submission\/35455","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/d3.harvard.edu\/platform-rctom\/wp-json\/wp\/v2\/hck-submission"}],"about":[{"href":"https:\/\/d3.harvard.edu\/platform-rctom\/wp-json\/wp\/v2\/types\/hck-submission"}],"author":[{"embeddable":true,"href":"https:\/\/d3.harvard.edu\/platform-rctom\/wp-json\/wp\/v2\/users\/11573"}],"replies":[{"embeddable":true,"href":"https:\/\/d3.harvard.edu\/platform-rctom\/wp-json\/wp\/v2\/comments?post=35455"}],"version-history":[{"count":0,"href":"https:\/\/d3.harvard.edu\/platform-rctom\/wp-json\/wp\/v2\/hck-submission\/35455\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/d3.harvard.edu\/platform-rctom\/wp-json\/wp\/v2\/media\/35456"}],"wp:attachment":[{"href":"https:\/\/d3.harvard.edu\/platform-rctom\/wp-json\/wp\/v2\/media?parent=35455"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/d3.harvard.edu\/platform-rctom\/wp-json\/wp\/v2\/categories?post=35455"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}