The second draft of my English Composition II research paper is turned in and getting graded. I’m still not sure how this story ends. As to the answer to the question in the title, Betteridge’s Law of Headlines suggests the answer is, “No.” Future revisions will explore why and what we can do about it. In the conclusion, I explore examples of too little protection, as well as too much, and how important it is that we get this right; because the training datasets for machine learning systems probably won’t rely on scooping up vast amounts of random user data for long.
“Copyright protection subsists, in accordance with this title, in original works of authorship fixed in any tangible medium of expression, now known or later developed, from which they can be perceived, reproduced, or otherwise communicated, either directly or with the aid of a machine or device.”(U.S. House of Representatives §106).
With the aid of a new type of machine, a learning machine, our digital future is being rewritten before our eyes. It is being generated, extruded, fabricated, hallucinated, and confabulated in ways we did not, and perhaps could not have foreseen.
Deep learning machines were theorized in the 1950s, but recent advances in computing, driven by computer gaming and cryptocurrencies, have accelerated machine learning systems in the last five years, powering devices and services we rely on daily at work, home, and play. (Littman et al. SQ2).
Put simply, deep learning systems are computer algorithms that simulate biological neurons, that are then trained on vast quantities of data, gathered from public and private data stores. These machine learning systems are capable of achieving near-miraculous feats of natural language translation; visual and audio recognition; near-perfect voice, music, and sound generation; and the creation of fantastic works of art and photography, including the generation of architectural designs of realistic and unrealistic nature, as well as realistic simulations of organic structures that make up the natural world.
Digital platforms that played host to thriving communities of artists and writers, once considered the sole domain of human creativity, have been overwhelmed with ersatz works that challenge our notions of creativity itself. And ownership.
The backlash came swiftly on the heels of DALL•E, Stable Diffusion, and Midjourney: Generative AI is a legal problem, to be decided by the courts. (Wiggers).
The US Copyright Office was quick to rule that they, “will not register works produced by a machine or mere mechanical process that operates randomly or automatically without any creative input or intervention from a human author.”
A vital open question remains, however: Copyright law was intended to encourage humans to share ideas, so that we might all benefit from them together, to enhance and benefit our commonwealth. Does the same apply to machines when they learn?
How does an idea become property?
Ideas are the raw material from which the arts and literature is shaped. We adorn them and develop them into something beautiful: like taking simple flour and humble eggs and turning them into an everything bagel, or even better yet, a chocolate cake.
Of course, anyone can go to the grocery store and buy whole wheat pastry flour, brown eggs, cocoa powder, brown sugar, salted butter, whole milk, and baking powder; but that does not make one a baker. A baker is a highly-trained combiner of those raw materials, with the skills to mix and combine them in a hundred different permutations, only one of which, when mixed well and poured into a greased cake pan, will culminate in the perfect layer of chocolate cake. Moist and delicious.
US Copyright Law, then, tries to protect the baker who bakes and sells delicious cakes, without interrupting people’s shopping trips with searches, seizures, and cake-ingredient-rights management. The constitution protects our right to eggs and butter for a nutritious breakfast omelette without government interference. That is to say, copyright law protects your right to make your cake, and my right to make my cake; but not the idea of cake, or even the recipe. (Library of Congress ArtI.S8.C8.1.1).
On the other side of the kitchen, the US Patent and Trademark Office attempts to protect the mechanical baking process, for instance, that makes hundreds of cakes a day to be sold on grocery store shelves. Those cakes have very specific recipes, and you can find examples of them all over the internet. Chef Claire Saffitz is noteworthy for her extravagant attempts to recreate mass-produced candy in the Bon Appétit test kitchens. US Patent Law protects the process of mass producing the candies, but does not prevent you or I from making exact replicas of them one or even twenty at a time, at home, for our own personal enjoyment, or for sharing with friends and family.
In fact, if we wanted to make a “Twix™”-like cookie, invent our own mass-production factory (using robots, of course), name those cookies something unique (“RoboCocoaWafers”, or some-such), so as not to run afoul of US Patent and Trademark Law, we could. We would be following in the tradition of Nabisco’s “Oreo™” cookies, as they replaced Sunshine Biscuits’ “Hydrox™” cookies in America’s cupboards and lunchboxes. Hydrox cookies were invented, baked, and sold years earlier; and the dispute between cookie-makers goes on to this day, over 100 years later. (Sales)
The recipe for each cookie may be similar, but in the case of “Secret Recipes”, the US Patent and Trademark Office has special protections that make up US Trade Secret Policy that protect your recipe from prying eyes, but do not prohibit other cookie makers from independent discovery of or reverse engineering your recipe. They only prevent outright theft when the offending party intends or knows “that the offense will, injure any owner of that trade secret.”
This Cookie War vividly illustrates how our system of protection for specific outcomes of the artistic and invention process is quite limited: even Vector design of a chocolate cake surrounded by its ingredients. Overlaid are icons representing copyright, patent, and trademark laws, with a playful twist.Vector design of a chocolate cake surrounded by its ingredients. Overlaid are icons representing copyright, patent, and trademark laws, with a playful twist.wholesome baked goods aren’t safe from copycats.
“In the age of machine learning and vast digital landscapes, the nuances of intellectual property become even more intricate. While a writer’s unique expression remains safeguarded, the core ideas they introduce flow freely, becoming threads in society’s ever-evolving tapestry. The balance lies in recognizing the value of individual creativity while appreciating the communal dance of ideas. Just as with cookies on a store shelf, the words penned by authors are out there for consumption, inspiration, and adaptation. The challenge is ensuring that original creators feel acknowledged and valued even as their ideas inspire new generations of thinkers.”Conversation with ChatGPT Plus
From the Bakery to the Library
Books and articles are not cookies, of course, nor even cakes. The recipe for arts and literature lists the sacrifice of no small amount of blood, sweat, and tears. Artists and writers pour their hearts and souls into their work, and rightly expect something in exchange for that. After all, brushes and paints; or computers and software, don’t come cheap. It is every starving artists’ dream that one day they might starve a little less frequently, or at least not to be left with an empty glass when the bottle of absinthe is passed around.
There is something ineffable about the written word and paint on canvas, or a well-framed photograph, or shaped clay and hewn stone. They seem uniquely human, because we value the effects they have on our lives almost as much as we value the emotional labor creators pour into their opuses. We build libraries and museums to laud the efforts and proudly display the works of our beloved scribes and artisans.
But few creators will be at the table when the billions and billions of US dollars are passed out by Silicon Valley venture capitalists in the digital gold rush for artificial intelligence, despite much of that money being used to scrape our portfolios for examples of our work to train the machine that will replace us.
The Congressional Research Service’s report to congress quotes the most well-known cookie-counterfeiter: OpenAI. As it defends itself from accusations that they have committed copyright violations in their training of ChatGPT, they state in their defense as reported by Christopher T. Zirpoli of the Congressional Research Service, “that ‘[w]ell-constructed AI systems generally do not regenerate, in any nontrivial portion, unaltered data from any particular work in their training corpus.’ Thus, OpenAI states, infringement ‘is an unlikely accidental outcome.’”
We stand at a crossroads. The well-lit reflective traffic sign for “Oreo Cookie Tollroad” hovers over a six-lane superhighway, next to a weaving dirt trail that leads off into into the wilderness, where a weather-worn, wooden signpost reads, “Hydrox Trailhead”.
There are dangers to overprotecting our cookie recipes. In a recent court case, copyright law was used to protect an author from harassment by members of the notorious Kiwi Farms hacker group. A member of the hacker group encouraged other members to freely share the author’s works in a way that would prevent them from earning money from its publication. The ruling, and the disagreement around it, mirrors in some ways the actions of anti-generative-AI activists; who believe that their only protection from having their livelihoods subsumed is copyright law, forcefully applied, such that the concept of “fair use” as outlined, is diminished. “We definitely don’t want more copyright doctrines that facilitate pernicious removals,” noted Technology & Marketing Law Blog publisher, Prof. Eric Goldman. (Belanger).
And yet, giving free reign over our very personal data to organizations ranging from the shadowy Kiwi Farms to well-funded corporations like Open AI, Google DeepMind, Stability AI, Midjourney Inc., Inflection AI, Anthropic, and an ever-growing list of “Surveillance Capitalists”, threatens to undermine democratic protections like the right to privacy. Social Psychologist Shoshana Zuboff makes clear the stakes are high: “Democracy is on the ropes in the UK, US, many other countries. Not in small measure because of the operations of surveillance capitalism.” (Kavenna).
Belanger, Ashley. “Kiwi Farms ruling sets “dubious” copyright precedent, expert warns”. Ars Technica, 18 Oct. 2023, https://arstechnica.com/tech-policy/2023/10/court-ruling-may-doom-kiwi-farms-but-set-dubious-copyright-precedent/. Accessed 22 Oct. 2023.
ChatGPT Plus. “Copyright Research” chat transcript. OpenAI, Generated 9 Oct. 2023, https://chat.openai.com/c/9ca663c6-48cc-4d78-9487-8abfa0e62791.
Kavenna, Joanna. “Shoshana Zuboff: Surveillance capitalism is an assault on human autonomy.” The Guardian, 4 Oct. 2019, https://www.theguardian.com/books/2019/oct/04/shoshana-zuboff-surveillance-capitalism-assault-human-automomy-digital-privacy. Accessed 23 Oct. 2023.
Littman, Michael L. et al. Gathering Strength, Gathering Storms: The One Hundred Year Study on Artificial Intelligence (AI100) 2021 Study Panel Report. Stanford University, Sept. 2021, http://ai100.stanford.edu/2021-report. Accessed 23 Oct. 2023.
Library of Congress. “ArtI.S8.C8.1.1 Origins and Scope of the Power.” Cornell Law School Legal Information Institute, https://www.law.cornell.edu/constitution-conan/article-1/section-8/clause-8/origins-and-scope-of-the-power. Accessed 9 Oct. 2023.
Saffitz, Claire. “Pastry Chef Attempts to Make Gourmet Twix.” Bon Appétit YouTube Channel, 13 Jun. 2019, https://www.youtube.com/watch?v=qOsp6IMqnn0. Accessed 9 Oct. 2023.
Sales, Ben. “Hydrox, Original Kosher Sandwich Cookie, Accuses Oreo of Sabotage.” The American Israelite, 15 Aug. 2018. https://sbctc-shoreline.primo.exlibrisgroup.com/permalink/01STATEWA_SHORLINE/6olkva/cdi_proquest_newspapers_2088748080. Accessed 9 Oct. 2023.
“Trade Secret Policy.” US Patent and Trademark Office, https://www.uspto.gov/ip-policy/trade-secret-policy. Accessed 23 Oct. 2023.
US Copyright Office. “Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence.” Copyright.gov, 16 Mar. 2023, https://copyright.gov/ai/ai_policy_guidance.pdf. Accessed 23 Oct. 2023.
Wiggers, Kyle. “The current legal cases against generative AI are just the beginning.” TechCrunch, 27 Jan. 2023, https://techcrunch.com/2023/01/27/the-current-legal-cases-against-generative-ai-are-just-the-beginning/. Accessed 23 Oct. 2023.
Zirpoli, Christopher T. “Generative Artificial Intelligence and Copyright Law.” Congressional Research Service, 29 Sept. 2023, https://crsreports.congress.gov/product/pdf/LSB/LSB10922. Accessed 9 Oct. 2023.
Image created by DALL-E 3. https://www.bing.com/image/create/ prompted thusly, “Vector design of a chocolate cake surrounded by its ingredients. Overlaid are icons representing copyright, patent, and trademark laws, with a playful twist.”