DALL-E Does Palsgraf

JOURNAL OF LAW, TECHNOLOGY & THE INTERNET • VOL. 14 • NO. 1 • 2022 – 2023 DALL-E DOES PALSGRAF Bryant Walker Smith* F0 What happens when we ask a leading artificial intelligence (AI) tool for image generation to illustrate the facts of a leading law school case? This article does just that. I first introduce this tool specifically and machine learning generally. I then summarize the seminal case of Palsgraf v. Long Island Railroad. For the main event, I show the images that the tool created based on the facts as the majority and dissent recount them. Finally, I translate this exercise into lessons for how lawyers and the law should think about AI. * Associate Professor of Law and (by courtesy) Engineering, University of South Carolina; Affiliate Scholar, Center for Internet and Society at Stanford Law School. Bryant’s publications are available at newlypossible.org. Chief Judge Cardozo, Judge Andrews, and the creators of DALL-E deserve most of the credit (and none of the blame) for this article. CONTENTS I. Introduction .......................................................................................................... 89 II. Dall-E, Machine Learning, And Image Generation .............................................. 90 III. Nondeterministic Systems .................................................................................... 93 IV. Palsgraf v. Long Island Railroad Co.................................................................... 95 V. Cardozo’s Facts .................................................................................................... 99 VI. Andrews’s Facts ................................................................................................. 110 VII. Some Human-Generated Thoughts .................................................................... 117 VIII. Conclusion: Cardozo Versus Andrews ............................................................... 123 IX. Postscript: ChatGPT ........................................................................................... 126 i Journal of Law, Technology & the Internet — Vol. 14 I. INTRODUCTION What happens when we ask a leading artificial intelligence (AI) tool for image generation to illustrate the facts of a leading law school case? This article does just that. I first introduce this tool specifically and machine learning generally. I then summarize the seminal case of Palsgraf v. Long Island Railroad. For the main event, I show the images that the tool created based on the facts as the majority and dissent recount them. Finally, I translate this exercise into lessons for how lawyers and the law should think about AI. 89 Journal of Law, Technology & the Internet — Vol. 14 II. DALL-E, MACHINE LEARNING, AND IMAGE GENERATION DALL-E is a computer tool that generates photorealistic images based on text supplied by the user. 1 For example, in response to the phrase “a fancy law school classroom with a cat professor,” DALL-E created these four original images: F1 DALL-E is developed and maintained by the OpenAI organization, which seeks “to ensure that artificial general intelligence (AGI)—by which we mean highly autonomous systems that outperform humans at most economically valuable work—benefits all of humanity.”2 DALL-E itself is an example of F2 1 DALL-E 2, OPEN AI, https://openai.com/dall-e-2/ (last visited Nov. 9, 2022) [https://perma.cc/E8ER-UT9C]. “DALL-E 2” has been rebranded just “DALL-E.” 2 Open AI Charter, OPEN AI, https://openai.com/charter (last visited Sept. 9, 2022) [https://perma.cc/42WC-WNGB]. 90 Dall-E Does Palsgraf specific artificial intelligence rather than AGI, which does not yet exist.3 When this article was written, access to DALL-E was by invitation only and conditioned on adherence to OpenAI’s content policy. 4 It is now available publicly. Similar tools are also available. 5 F3 F4 F5 In general, tools for recognizing or generating images initially learn by processing huge numbers of images that are each linked in some way with descriptive text. 6 Through this training, these artificial neural networks develop relationships between various visual and textual elements to create what is in effect a much more complex and dynamic version of a thesaurus. 7 F6 F7 There are a range of approaches to training these neural networks. In a traditional model of supervised learning, humans manually label each image in the training dataset. One popular dataset, ImageNet, “required over 25,000 workers to annotate 14 million images for 22,000 object categories.” 8 Amazon’s F8 3 Will Douglas Heaven, Artificial General Intelligence: Are We Close, and Does It Even Make Sense to Try?, MIT TECH. REV. (Oct. 15, 2020), https://www.technologyreview.com/2020/10/15/1010461/artificial-general-intelligence-robots-aiagi-deepmind-google-openai/ [https://perma.cc/3LE9-R5SK]; Gary Marcus, Artificial General Intelligence Is Not as Imminent as You Might Think, SCI. AM. (July 1, 2022), https://www.scientificamerican.com/article/artificial-general-intelligence-is-not-as-imminent-asyou-might-think1/ [https://perma.cc/QX46-JFLN]. 4 Content Policy, DALL-E (July 20, 2022), https://labs.openai.com/policies/content-policy [https://perma.cc/G9D3-5B7A]. 5 See, e.g., MIDJOURNEY, https://www.midjourney.com (last visited Sept. 9, 2022) [https://perma.cc/CY2J-MZPA]; Emad Mostaque, Stable Diffusion Public Release, STABILITY.AI (Aug. 22, 2022) [https://perma.cc/Y849-VQXN], https://stability.ai/blog/stable-diffusion-publicrelease; AI Art Generator, NIGHTCAFÉ, https://nightcafe.studio (last visited Sept. 9, 2022) [https://perma.cc/BFK4-M4FH]; Create Art with AI, STARRYAI, https://www.starryai.com (last visited Sept. 9, 2022) [https://perma.cc/K585-W9BJ]; Online Photo Editor for Everyone, FOTOR, https://www.fotor.com/ (last visited Sept. 9, 2022) [https://perma.cc/3KE5-CRU3]. Craiyon is related to DALL-E 2’s predecessor. CRAIYON, https://www.craiyon.com (last visited Sept. 9, 2022) [https://perma.cc/6CNL-K6FJ]. 6 See, e.g., Hannah Murdock, What is DALL-E Mini? How an AI image generator is making the internet’s weirdest memes, DESERTNEWS (Aug. 13, 2022) https://www.deseret.com/2022/8/13/23207472/dall-e-mini-ai-image-generator-craiyon-how-touse-machine-learning-how-does-it-work [https://perma.cc/9TYV-ZXK5]. 7 A thesaurus might treat “boat” as very similar to “ship,” somewhat similar to “vehicle,” and not at all similar to “alfalfa.” A neural network designed and trained for image recognition might treat visual elements that correspond with a hull’s interface with water as strongly associated with “boat,” weakly associated with “vehicle,” and not at all associated with “alfalfa.” 8 Alec Radford et al., CLIP: Connecting Text and Images, OPEN AI (Jan. 5, 2021), https://openai.com/blog/clip/ [https://perma.cc/54M8-U3UC]. 91 Journal of Law, Technology & the Internet — Vol. 14 Mechanical Turk is an example of a platform that connects developers with workers who ar (...truncated)