You Look Like a Thing and I Love You

by Janelle Shane

  Thanks to my friends and family, who encouraged me during this long process, who listened to my practice talks and laughed at my jokes, and who were always ready to help me recharge with some tunes, some hiking, or some culinary experiments.

  And finally, thanks to all my readers and followers at aiweirdness.com, who have already made so much of my strange AI experiments into reality—the knitting patterns, the cookies, the nail polish, the burlesque shows, the weird creatures, the absurd cat names, the beer names, and even the opera. Look what we made now! May the giraffes be ever with you.

  Discover Your Next Great Read

  Get sneak peeks, book recommendations, and news about your favorite authors.

  Tap here to learn more.

  About the Author

  Janelle Shane has a PhD in electrical engineering and a masters in physics. At aiweirdness.com, she writes about artificial intelligence and the hilarious and sometimes unsettling ways that algorithms get human things wrong. She was named one of Fast Company’s 100 Most Creative People in Business and is a 2019 TED Talks speaker. Her work has appeared in the New York Times, Slate, The New Yorker, The Atlantic, Popular Science, and more. She is almost certainly not a robot.

  Notes

  Introduction

  1. Caroline O’Donovan et al., “We Followed YouTube’s Recommendation Algorithm Down the Rabbit Hole,” BuzzFeed News, January 24, 2019, https://www.buzzfeednews.com/article/carolineodonovan/down-youtubes-recommendation-rabbithole.

  Chapter 1: What is AI?

  1. Joel Lehman et al., “The Surprising Creativity of Digital Evolution: A Collection of Anecdotes from the Evolutionary Computation and Artificial Life Research Communities,” ArXiv:1803.03453 [Cs], March 9, 2018, http://arxiv.org/abs/1803.03453.

  2. Neel V. Patel, “Why Doctors Aren’t Afraid of Better, More Efficient AI Diagnosing Cancer,” The Daily Beast, December 11, 2017, https://www.thedailybeast.com/why-doctors-arent-afraid-of-better-more-efficient-ai-diagnosing-cancer.

  3. Jeff Larson et al., “How We Analyzed the COMPAS Recidivism Algorithm,” ProPublica, May 23, 2016, https://www.propublica.org/article/how-we-analyzed-the-compas-recidivism-algorithm.

  4. Chris Williams, “AI Guru Ng: Fearing a Rise of Killer Robots Is Like Worrying about Overpopulation on Mars,” The Register, March 19, 2015, https://www.theregister.co.uk/2015/03/19/andrew_ng_baidu_ai/.

  5. Marianne Bertrand and Sendhil Mullainathan, “Are Emily and Greg More Employable Than Lakisha and Jamal? A Field Experiment on Labor Market Discrimination,” American Economic Review 94, no. 4 (September 2004): 991–1013, https://doi.org/10.1257/0002828042002561.

  Chapter 2: AI is everywhere, but where is it exactly?

  1. Stephen Chen, “A Giant Farm in China Is Breeding 6 Billion Cockroaches a Year. Here’s Why,” South China Morning Post, April 19, 2018, https://www.scmp.com/news/china/society/article/2142316/giant-indoor-farm-china-breeding-six-billion-cockroaches-year.

  2. Heliograf, “High School Football This Week: Einstein at Quince Orchard,” Washington Post, October 13, 2017, https://www.washingtonpost.com/allmetsports/2017-fall/games/football/87408/.

  3. Li L’Estrade, “MittMedia Homeowners Bot Boosts Digital Subscriptions with Automated Articles,” International News Media Association (INMA), June 18, 2018, https://www.inma.org/blogs/ideas/post.cfm/mittmedia-homeowners-bot-boosts-digital-subscriptions-with-automated-articles.

  4. Jaclyn Peiser, “The Rise of the Robot Reporter,” New York Times, February 5, 2019, https://www.nytimes.com/2019/02/05/business/media/artificial-intelligence-journalism-robots.html.

  5. Christopher J. Shallue and Andrew Vanderburg, “Identifying Exoplanets with Deep Learning: A Five Planet Resonant Chain around Kepler-80 and an Eighth Planet around Kepler-90,” The Astronomical Journal 155, no. 2 (January 30, 2018): 94, https://doi.org/10.3847/1538-3881/aa9e09.

  6. R. Benton Metcalf et al., “The Strong Gravitational Lens Finding Challenge,” Astronomy & Astrophysics 625 (May 2019): A119, https://doi.org/10.1051/0004-6361/201832797.

  7. Avi Bagla, “#StarringJohnCho Level 2: Using DeepFakes for Representation,” YouTube video, posted April 9, 2018, https://www.youtube.com/watch?v=hlZkATlqDSM&feature=youtu.be.

  8. Tom Simonite, “Facebook Built the Perfect Chatbot but Can’t Give It to You Yet,” MIT Technology Review, April 14, 2017, https://www.technologyreview.com/s/604117/facebooks-perfect-impossible-chatbot/.

  9. Ibid.

  10. Casey Newton, “Facebook Is Shutting Down M, Its Personal Assistant Service That Combined Humans and AI,” The Verge, January 8, 2018, https://www.theverge.com/2018/1/8/16856654/facebook-m-shutdown-bots-ai.

  11. Andrew J. Hawkins, “Inside Waymo’s Strategy to Grow the Best Brains for Self-Driving Cars,” The Verge, May 9, 2018, https://www.theverge.com/2018/5/9/17307156/google-waymo-driverless-cars-deep-learning-neural-net-interview.

  12. “OpenAI Five,” OpenAI, accessed August 3, 2019, https://openai.com/five/.

  13. Katyanna Quatch, “OpenAI Bots Smashed in Their First Clash against Human Dota 2 Pros,” The Register, August 23, 2018, https://www.theregister.co.uk/2018/08/23/openai_bots_defeated/.

  14. Tom Murphy (@tom7), Twitter, August 23, 2018, https://twitter.com/tom7/status/1032756005107580929.

  15. Mike Cook (@mtrc), Twitter, August 23, 2018, https://twitter.com/mtrc/status/1032783369254432773.

  16. Tom Murphy, “The First Level of Super Mario Bros. Is Easy with Lexicographic Orderings and Time Travel… After That It Gets a Little Tricky” (research paper, Carnegie Melon University), April 1, 2013, http://www.cs.cmu.edu/~tom7/mario/mario.pdf.

  17. Benjamin Solnik et al., “Bayesian Optimization for a Better Dessert” (paper presented at the 2017 NIPS Workshop on Bayesian Optimization, Long Beach, CA, December 9, 2017), https://bayesopt.github.io/papers/2017/37.pdf.

  18. Sarah Kimmorley, “We Tasted the ‘Perfect’ Cookie Google Took 2 Months and 59 Batches to Create—and It Was Terrible,” Business Insider Australia, May 31, 2018, https://www.businessinsider.com.au/google-smart-cookie-ai-recipe-2018-5.

  19. Andrew Krok, “Waymo’s Self-Driving Cars Are Far from Perfect, Report Says,” Roadshow, August 28, 2018, https://www.cnet.com/roadshow/news/waymo-alleged-tech-troubles-report/.

  20. C. Lv et al., “Analysis of Autopilot Disengagements Occurring during Autonomous Vehicle Testing,” IEEE/CAA Journal of Automatica Sinica 5, no. 1 (January 2018): 58–68, https://doi.org/10.1109/JAS.2017.7510745.

  21. Andrew Krok, “Uber Self-Driving Car Saw Pedestrian 6 Seconds before Crash, NTSB Says,” Roadshow, May 24, 2018, https://www.cnet.com/roadshow/news/uber-self-driving-car-ntsb-preliminary-report/.

  22. Fred Lambert, “Tesla Elaborates on Autopilot’s Automatic Emergency Braking Capacity over Mobileye’s System,” Electrek (blog), July 2, 2016, https://electrek.co/2016/07/02/tesla-autopilot-mobileye-automatic-emergency-braking/.

  23. Naaman Zhou, “Volvo Admits Its Self-Driving Cars Are Confused by Kangaroos,” The Guardian, July 1, 2017, https://www.theguardian.com/technology/2017/jul/01/volvo-admits-its-self-driving-cars-are-confused-by-kangaroos.

  Chapter 3: How does it actually learn?

  1. Ian Goodfellow, Yoshua Bengio, and Aaron Courville, Deep Learning (Cambridge, Massachusetts: The MIT Press, 2016).

  2. Sean McGregor et al., “FlareNet: A Deep Learning Framework for Solar Phenomena Prediction” (paper presented at the 31st Conference on Neural Information Processing Systems, Long Beach, CA, December 8, 2017), https://dl4physicalsciences.github.io/files/nips_dlps_2017_5.pdf.

  3. Alec Radford, Rafal Jozefowicz, and Ilya Sutskever, “Learning to Generate Reviews and Discovering Sentiment,” ArXiv:1704.01444 [Cs], April 5, 2017, http://arxiv.org/abs/1704.01444.

  4. Andrej Karpathy, “The Unreasonable Effectiveness of Recurrent Neural Networks,” Andrej Karpathy Blog, May 21, 2015, http://karpathy.github.io/2015/05/21/rnn-effectiveness/.

  5. Chris Olah et al., “The Building Blocks of Interpretability,” Distill 3, no. 3 (March 6
, 2018): e10, https://doi.org/10.23915/distill.00010.

  6. David Bau et al., “GAN Dissection: Visualizing and Understanding Generative Adversarial Networks” (paper presented at the International Conference on Learning Representations, May 6–9, 2019), https://gandissect.csail.mit.edu/.

  7. “Botnik Apps,” Botnik, accessed August 3, 2019, ttp://botnik.org/apps.

  8. Paris Martineau, “Why Google Docs Is Gaslighting Everyone about Spelling: An Investigation,” The Outline, May 7, 2018, https://theoutline.com/post/4437/why-google-docs-thinks-real-words-are-misspelled.

  9. Shaokang Zhang et al., “Zoonotic Source Attribution of Salmonella enterica Serotype Typhimurium Using Genomic Surveillance Data, United States,” Emerging Infectious Diseases 25, no. 1 (2019): 82–91, https://doi.org/10.3201/eid2501.180835.

  10. Ian J. Goodfellow et al., “Generative Adversarial Networks,” ArXiv:1406.2661 [Cs, Stat], June 10, 2014, http://arxiv.org/abs/1406.2661.

  11. Ahmed Elgammal et al., “CAN: Creative Adversarial Networks, Generating ‘Art’ by Learning About Styles and Deviating from Style Norms,” ArXiv:1706.07068 [Cs], June 21, 2017, http://arxiv.org/abs/1706.07068.

  12. Beckett Mufson, “This Artist Is Teaching Neural Networks to Make Abstract Art,” Vice, May 22, 2016, https://www.vice.com/en_us/article/yp59mg/neural-network-abstract-machine-paintings.

  13. David Ha and Jürgen Schmidhuber, “World Models,” Zenodo, March 28, 2018, https://doi.org/10.5281/zenodo.1207631.

  Chapter 4: It’s trying!

  1. Tero Karras, Samuli Laine, and Timo Aila, “A Style-Based Generator Architecture for Generative Adversarial Networks,” ArXiv:1812.04948 [Cs, Stat], December 12, 2018, http://arxiv.org/abs/1812.04948.

  2. Emily Dreyfuss, “A Bot Panic Hits Amazon Mechanical Turk,” Wired, August 17, 2018, https://www.wired.com/story/amazon-mechanical-turk-bot-panic/.

  3. “COCO Dataset,” COCO: Common Objects in Context, http://cocodataset.org/#download. Images used during training were 2014 training + 2014 val, for a total of 124k images. Each dialog had 10 questions. https://visualdialog.org/data says 364m dialogs in the training set, so each image was encountered 364/1.24 = 293.5 times.

  4. Hawkins, “Inside Waymo’s Strategy.”

  5. Tero Karras et al., “Progressive Growing of GANs for Improved Quality, Stability, and Variation,” ArXiv:1710.10196 [Cs, Stat], October 27, 2017, http://arxiv.org/abs/1710.10196.

  6. Karras, Laine, and Aila, “A Style-Based Generator Architecture.”

  7. Melissa Eliott (0xabad1dea), “How Math Can Be Racist: Giraffing,” Tumblr, January 31, 2019, https://abad1dea.tumblr.com/post/182455506350/how-math-can-be-racist-giraffing.

  8. Corinne Purtill and Zoë Schlanger, “Wikipedia Rejected an Entry on a Nobel Prize Winner Because She Wasn’t Famous Enough,” Quartz, October 2, 2018, https://qz.com/1410909/wikipedia-had-rejected-nobel-prize-winner-donna-strickland-because-she-wasnt-famous-enough/.

  9. Jon Christian, “Why Is Google Translate Spitting Out Sinister Religious Prophecies?” Vice, July 20, 2018, https://www.vice.com/en_us/article/j5npeg/why-is-google-translate-spitting-out-sinister-religious-prophecies.

  10. Nicholas Carlini et al., “The Secret Sharer: Evaluating and Testing Unintended Memorization in Neural Networks,” ArXiv:1802.08232 [Cs], February 22, 2018, http://arxiv.org/abs/1802.08232.

  11. Jonas Jongejan et al., “Quick, Draw! The Data” (dataset for online game Quick, Draw!), accessed August 3, 2019, https://quickdraw.withgoogle.com/data.

  12. Jon Englesman (@engelsjk), Google AI Quickdraw Visualizer (web demo), Github, accessed August 3, 2019, https://engelsjk.github.io/web-demo-quickdraw-visualizer/.

  13. Gretchen McCulloch, “Autocomplete Presents the Best Version of You,” Wired, February 11, 2019, https://www.wired.com/story/autocomplete-presents-the-best-version-of-you/.

  14. Abhishek Das et al., “Visual Dialog,” ArXiv:1611.08669 [Cs], November 26, 2016, http://arxiv.org/abs/1611.08669.

  Chapter 5: What are you really asking for?

  1. @citizen_of_now, Twitter, March 15, 2018, https://twitter.com/citizen_of_now/status/974344339815129089.

  2. Doug Blank (@DougBlank), Twitter, April 13, 2018, https://twitter.com/DougBlank/status/984811881050329099.

  3. @Smingleigh, Twitter, November 7, 2018, https://twitter.com/Smingleigh/status/1060325665671692288.

  4. Christine Barron, “Pass the Butter // Pancake Bot,” Unity Connect, January 2018, https://connect.unity.com/p/pancake-bot.

  5. Alex Irpan, “Deep Reinforcement Learning Doesn’t Work Yet,” Sorta Insightful (blog), February 14, 2018, https://www.alexirpan.com/2018/02/14/rl-hard.html.

  6. Sterling Crispin (@sterlingcrispin), Twitter, April 16, 2018, https://twitter.com/sterlingcrispin/status/985967636302327808.

  7. Sara Chodosh, “The Problem with Cancer-Sniffing Dogs,” October 4, 2016, Popular Science, https://www.popsci.com/problem-with-cancer-sniffing-dogs/.

  8. Wikipedia, s.v. “Anti-Tank Dog,” last updated June 29, 2019, https://en.wikipedia.org/w/index.php?title=Anti-tank_dog&oldid=904053260.

  9. Anuschka de Rohan, “Why Dolphins Are Deep Thinkers,” The Guardian, July 3, 2003, https://www.theguardian.com/science/2003/jul/03/research.science.

  10. Sandeep Jauhar, “When Doctor’s Slam the Door,” New York Times Magazine, March 16, 2003, https://www.nytimes.com/2003/03/16/magazine/when-doctor-s-slam-the-door.html.

  11. Joel Rubin (@joelrubin), Twitter, December 6, 2017, https://twitter.com/joelrubin/status/938574971852304384.

  12. Joel Simon, “Evolving Floorplans,” joelsimon.net, accessed August 3, 2019, http://www.joelsimon.net/evo_floorplans.html.

  13. Murphy, “First Level of Super Mario Bros.”

  14. Tom Murphy (suckerpinch), “Computer Program that Learns to Play Classic NES Games,” YouTube video, posted April 1, 2013, https://www.youtube.com/watch?v=xOCurBYI_gY.

  15. Murphy, “First Level of Super Mario Bros.”

  16. Jack Clark and Dario Amodei, “Faulty Reward Functions in the Wild,” OpenAI, December 22, 2016, https://openai.com/blog/faulty-reward-functions/.

  17. Bitmob, “Dimming the Radiant AI in Oblivion,” VentureBeat (blog), December 17, 2010, https://venturebeat.com/2010/12/17/dimming-the-radiant-ai-in-oblivion/.

  18. cliffracer333, “So what happened to Oblivion’s npc ‘goal’ system that they used in the beta of the game. Is there a mod or a way to enable it again?” Reddit thread, June 10, 2016, https://www.reddit.com/r/oblivion/comments/4nimvh/so_what_happened_to_oblivions_npc_goal_system/.

  19. Sindya N. Bhanoo, “A Desert Spider with Astonishing Moves,” New York Times, May 4, 2014, https://www.nytimes.com/2014/05/06/science/a-desert-spider-with-astonishing-moves.html.

  20. Lehman et al., “The Surprising Creativity of Digital Evolution.”

  21. Jette Randløv and Preben Alstrøm, “Learning to Drive a Bicycle Using Reinforcement Learning and Shaping,” Proceedings of the Fifteenth International Conference on Machine Learning, ICML ’98 (San Francisco, CA: Morgan Kaufmann Publishers Inc., 1998), 463–471, http://dl.acm.org/citation.cfm?id=645527.757766.

  22. Yuval Tassa et al., “DeepMind Control Suite,” ArXiv:1801.00690 [Cs], January 2, 2018, http://arxiv.org/abs/1801.00690.

  23. Benjamin Recht, “Clues for Which I Search and Choose,” arg min blog, March 20, 2018, http://benjamin-recht.github.io/2018/03/20/mujocoloco/.

  24. @citizen_of_now, Twitter, March 15, 2018, https://twitter.com/citizen_of_now/status/974344339815129089.

  25. Westley Weimer, “Advances in Automated Program Repair and a Call to Arms,” Search Based Software Engineering, ed. Günther Ruhe and Yuanyuan Zhang (Berlin and Heidelberg: Springer, 2013), 1–3.

  26. Lehman et al., “The Surprising Creativity of Digital Evolution.”

  27. Yuri Burda et al., “Large-Scale Study of Curiosity-Driven Learning,” ArXiv:1808.04355 [Cs, Stat], August 13, 2018, http://arxiv.org/abs/1808.04355.

  28. A. Baranes and P.-Y. Oudeyer, “R-IAC: Robust Intrinsical
ly Motivated Exploration and Active Learning,” IEEE Transactions on Autonomous Mental Development 1, no. 3 (October 2009): 155–69, https://doi.org/10.1109/TAMD.2009.2037513.

  29. Devin Coldewey, “This Clever AI Hid Data from Its Creators to Cheat at Its Appointed Task,” TechCrunch, December 31, 2018, http://social.techcrunch.com/2018/12/31/this-clever-ai-hid-data-from-its-creators-to-cheat-at-its-appointed-task/.

  30. “YouTube Now: Why We Focus on Watch Time,” YouTube Creator Blog, August 10, 2012, https://youtube-creators.googleblog.com/2012/08/youtube-now-why-we-focus-on-watch-time.html.

  31. Guillaume Chaslot (@gchaslot), Twitter, February 9, 2019, https://twitter.com/gchaslot/status/1094359568052817920?s=21.

  32. “Continuing Our Work to Improve Recommendations on YouTube,” Official YouTube Blog, January 25, 2019, https://youtube.googleblog.com/2019/01/continuing-our-work-to-improve.html.

  Chapter 6: Hacking the Matrix, or AI finds a way

  1. Doug Blank (@DougBlank), Twitter, March 15, 2018, https://twitter.com/DougBlank/status/974244645214588930.

  2. Nick Stenning (@nickstenning), Twitter, April 9, 2018, https://twitter.com/DougBlank/status/974244645214588930

  3. Christian Gagné et al., “Human-Competitive Lens System Design with Evolution Strategies,” Applied Soft Computing 8, no. 4 (September 1, 2008): 1439–52, https://doi.org/10.1016/j.asoc.2007.10.018.

  4. Lehman et al., “The Surprising Creativity of Digital Evolution.”

  5. Karl Sims, “Evolving 3D Morphology and Behavior by Competition,” Artificial Life 1, no. 4 (July 1, 1994): 353–72, https://doi.org/10.1162/artl.1994.1.4.353.

  6. Karl Sims, “Evolving Virtual Creatures,” Proceedings of the 21st Annual Conference on Computer Graphics and Interactive Techniques, SIGGRAPH ’94 (New York: ACM, 1994), 15–22, https://doi.org/10.1145/192161.192167.