Image: Shutterstock
There has been lots of ink spilled over the issue of Artificial Intelligence (AI) developers free riding on the backs of creators by copying and ingesting their content without permission to produce AI “created” content, output that sometimes competes directly in the market with the original work. The most recent high-profile case involving this process is the complaint brought by the New York Times against OpenAI and indirectly Microsoft. Last week, I wrote about that case here. (When Giants Wrestle, the Earth Moves: NYT v OpenAI/Microsoft).
AI developers defend themselves by arguing, in the US context, that their actions constitute fair use. They claim that what they’re doing is no different from, say, an art student studying works by the Old Masters, imbibing inspiration from them, and then producing their own original work. The result of the AI process, they argue, is the same. It is creating something new through a “transformative” process such as the one that happens when a creator (artist, writer, composer, or other creator) takes inspiration from early works but then produces something new and original. The new work constitutes their own expression of the idea of which the earlier work was but a representation. However, there is a fundamental flaw with this argument when it comes to AI.
The argument presupposes that all creativity is merely a new combination of existing knowledge. Is creativity merely a rearrangement of what came before or is there an additional X factor introduced when something new is created? Does this X factor exist in AI-created works, or is AI exclusively based on the corpus of existing knowledge? Judging by the way it is trained, one must conclude it’s the latter. The algorithm is fed vast amounts of existing content that it regurgitates in response to prompts. Sometimes the inputs are not evident, especially if there are vast numbers of them; sometimes they are all too evident as in the case of ChatGPT producing virtually verbatim excerpts from the New York Times. But in neither case is it apparent to me that anything truly “new” or “creative” has been produced. An AI algorithm is not a proxy for the human brain or for human creation.
Even with vast numbers of inputs scraped from the internet—the good, bad and the ugly—an AI machine can only produce what it has been exposed to. Sometimes it produces a mish-mashed version of what it has been fed, where individual inputs are not obvious. At other times it resorts to total plagiarism, and at yet other times it produces a pale and shallow imitation of the original. So why not go to the original? Moreover, the conclusions AI draws from the data it has been trained on can be, and often are, completely wrong from a factual point of view. The NYT complaint outlined convincingly how AI can just make up answers from its database when it doesn’t have the facts. I suppose you could say this shows an element of “creativity”, but it is creativity of a blind and completely irresponsible kind.
In reflecting on this, I am thinking of my own modest efforts at creativity, particularly my recent book, “In Defence of Copyright”. As an AI machine would do, I consulted a lot of sources (but did not copy their work other than to register the content in my brain and make a few notes where I wanted to cite something specific.) I then stitched together a narrative that drew on a lot of existing knowledge, with proper citations and credit of course. However, along the way, as part of the writing process, elements of personal creativity inevitably crept in. Personal experiences that only I have had influenced my thinking and conclusions. No one else could have produced this work in exactly the same way, even if I had given them all my reference materials. The shaping of the book came from me. I decided what to put in and what to leave out. I decided the order of the presentation. I decided how various points would be expressed, what the tone would be, and what particular terms and references to use. And I drew the conclusions in the book. Could an AI “machine” have done this?
If I had put in the prompts, “write me a book explaining what copyright is and how it works, why it’s important, covering its history and current challenges, providing some examples of controversies in copyright law along with some case studies and entertaining anecdotes”, AI would have produced a generic version but I doubt that the work would bear much similarity to my expression of these ideas. It would have a pastiche of ideas of others but without the X factor of personal interpretation.
If you want something “new” and truly creative, with new insights and elements, don’t rely on AI. It can’t create what it has not been fed, although it can get the content wrong. AI has no true creative capabilities; in my view it is simply an artificial reorganization of existing ideas.
And while I am on the topic of creativity, even though I swore that unlike in the past I would not be tempted this year to comment on the annual hyperbole about works coming into the public domain, like Disney’s Steamboat Willie, I am going to succumb. Willie has managed to capture the usual facile press attention devoted to the annual crop of works that fall into the public domain in the US each January 1. I was provoked to make these comments by an interview on CBC radio in which a law professor at UBC, after providing a balanced and thoughtful overview of what Willie’s entry into the public domain in the US means in Canada, and how those wishing to use Willie’s image to create new works need to be careful, then waxed on about how meaningful this was because it would unleash a flood of creative derivative works based on the 1928 cartoon version of Mickey. A horror version, where Mickey as Willie is depicted as a violent slasher, is one of the first out of the gate It is directed, I am ashamed to say, by a Canadian. Variety has a story about another equally inane production, this one American, in which Mickey torments a group of unsuspecting ferry passengers. How has the world been able to survive without these creations until now?
Our professor, who is a comic book fan, enthused about how exciting it will be when Superman enters the public domain. Superman derivatives will then be freely produced by all these people waiting to clip his wings. Superman was first published by DC Comics in 1938 so it will be a while before his image will be defaced by “creators” seeking to, in effect, free ride on the reputation of a successful character who has been developed over the years by the rights-holder. Of course, the idea of a flying human with supernatural powers cannot be copyrighted. It is the expression of that particular idea, in this case in the personality of Superman, that is protected. The existence of a range of other superheroes, including some who can fly, such as Captain Marvel and Green Lantern, is ample proof of this. So why do all these creative people, who now have access to all kinds of technology to produce content attractive to consumers, need to exploit an existing character, and make it do weird and out-of-character things? Because they are not truly creative!
A true creator would invent a new character, not exploit one where all the heavy lifting of building name recognition has been done by someone else. Successful writers may be inspired by the works of other authors, but they don’t wait until those works fall into the public domain and hijack the characters and reputation of the work to produce a warmed-over version of the original. I can’t understand why supposedly creative people are so obsessed by well-known works falling into the public domain, especially if they have the name Disney associated with them. Is it because they need the crutch of the Disney reputation to get anyone to pay attention to works that hardly matter—and lack much creativity?
Anything for a quick buck, I guess, using the veneer of “creativity” to justify what is really a paucity of imagination. AI is not all that different. It can produce content that is “good enough” for some routine and mundane purposes, but in doing so it risks stifling true creativity. AI needs to be harnessed for the greater good, not allowed to become the dominant producer of content, in the process threatening to replace true and original creation. Once the well of human creation dries up, there will be no water to lubricate the AI machine. We need to manage AI carefully or face the consequences of a cultural drought.
© Hugh Stephens, 2024
