First mover disadvantage: generative AI, IP theft and the origin of ideas
The other day, I was scrolling through my LinkedIn feed and couldn’t help but notice the growing prevalence of unoriginal content. Images, videos and posts originally gracing the platform months and years ago, repackaged by somebody without reference to the original creator. It’s smart when you think about it. Having knowledge of the post’s prior performance, removing the most time-consuming part of the process (composing original content), and removing the second most time-consuming part of the process (identifying the creator), the influencer can attain a predictable level of engagement with little-to-no effort. The risk? Well, somebody might call you out, as James Corden, Converse, Jay Shetty, et al. have learned in recent years.
According to Jonathan Bailey of Plagiarism Today (see what I did there 😉), jokes “enjoy very limited legal protection”. This is apparently due to the Idea/Expression Dichotomy, protecting the expression of a joke but not the idea behind it. The result? The emergence, over the past half-century, of social norms that threaten to reputationally end the career of plagiarists. Power and influence are interesting factors to account for here, as Joe Rogan famously learned in the mid-2000s when calling out influential comedian-cum-joke-thief, Carlos Mencia.
…but what if the alleged IP thief isn’t human?
In November 2022, worldwide online search interest for a three-letter acronym (GPT) barely registered. Since then, interest has skyrocketed — if you thought the boom in metaverse interest 12 months ago was big, it is dwarfed in comparison (see chart below). GPT stands for Generative Pre-trained Transformer. GPT-3.5 is one of the largest and most advanced language models currently available. ChatGPT, released by OpenAI in November 2022, essentially works as a chatbot, enabling users to ask questions and follow-up questions. Since its launch, users have tested the technology with university exam questions, poem requests, jokes and computer code. So much for the learn-to-code movement!
Other creativity-emulating platforms garnering recent attention include Midjourney and Lensa — which enable users to generate visual artwork from a description or have their own images retouched in a certain artistic style. Classified together, these platforms are termed “generative AI”. Although many of these platforms are new and exciting, generative AI has attracted criticism. Artists have claimed that the sources used to train these models are in fact their own human creations — giving rise to the question of how reputational risk for IP theft works when the alleged thief isn’t human.
“There is no such thing as a new idea…”
Who was first: Uber or Lyft? Well, according to PitchBook, Lyft was founded in 2007 (the year of the iPhone) and Uber in 2009. But what about the BlaBlaCar carpool platform founded in 2006? Or the famed London taxi of the 20th century? Or the Elizabethan-era horse-drawn Hackney Coaches? What of the billions in public funds that went toward roads and highways to make it all possible? The smartphone designers and manufacturers? Can we really credit a single individual or innovation for the privileges of ridesharing we enjoy today? According to PitchBook, there have been 712 ridesharing companies founded since 2005 (see chart below). In 2007 when Lyft was founded, so too were 12 others.
A century before the founding of Lyft, Mark Twain claimed there was “no such thing as a new idea…” Ironically, he too was beaten to the punch on this claim— by more than two millennia:
What has been will be again, what has been done will be done again; there is nothing new under the sun
Ecclesiastes 1:9
Defensibility and first mover disadvantage
As with plagiarised LinkedIn posts, so too with entrepreneurship. Since at least the 1980s, marketing and innovation literature has known of the potential benefits of being a late mover. A 1998 paper by Shankar, Carpenter and Krishnamurthi shared a minimum of two ways late movers can outsell first movers or pioneers. First, “a late mover can beat a pioneer at the pioneer’s own game” — in particular by allowing the pioneer to do all the heavy lifting in defining the category, educating consumers, setting prices, etc. Second, a late mover can overtake a pioneer through innovation (arguably what ride sharing apps did to motorised taxis, and what motorised taxis did to horse-drawn cabs). Defensibility is so important for this reason, as explained by Pi Labs Principal, Hugo Silva:
One of the main reasons we look for product defensibility is because of its impact on user stickiness. If your product has something competitors don’t, consumers will likely pay more for this value to be returned to them. For example, our portfolio company Contilio isn’t alone in its category, but a combination of inputs (technologies being applied), process speed, as well as output quality and variety, offers its customers savings on costs and time consumed which is difficult for competitors to replicate to the same extent.
Beat them or join them..?
After a recent conversation with Alicia Navarro, founder of FLOWN, I learned of her surprise at how quickly her team and users had adopted ChatGPT following its launch. Alicia expressed cautious optimism with the technology, citing the new superskill of the year as “knowing what questions to ask it, and how to effectively make use of the outputs”. Widespread adoption of generative AI would have some predictable outcomes “education will need to change, as will the way students are assessed” she said. This perspective among entrepreneurs seems to be growing in prevalence. Gaurav Bajaj, co-founder of Little Place Labs, recently said “AI will not take your job, but a person who can keep up with the growing versions of AI will” — in reference to the upcoming launch of GPT-4, which purportedly increases the parameter quantity to 100 trillion (up from 175 billion with GPT-3).
Lessons from a little Scottish town…
History teaches a valuable lesson comparable to the above perspectives. It applies to a textile motif some estimate to date back thousands of years. I am speaking, of course, of paisley — named after the Scottish town from where it by-no-means originates. As the story goes, the motif initially referred to as “boteh” or “buto” was derived from the Zoroastrian symbol of a cypress tree — hand woven from fine goat hair into shawls. It became a coveted garment revered by Persian royalty, and eventually 19th century European high society (most notably Empress Josephine). Britain was in the midst of the First Industrial Revolution at this time, meaning popular demand for the fashion of high society could be met at a more accessible price. The Scottish town of Paisley had become a centre for textile manufacture, and therefore served as the perfect place to locally produce paisley en masse. With production levels enabling prices accessible to those wretched commoners, paisley became uncool for the social elite for the first time. Production levels also caused artisans, who produced the motif by hand, to reach a crossroads: either rage against the machine or become part of it. It appears few pursued the latter option (or more likely, were unable to), leading to British industrial imitators of the design being “largely devoid of any artistic excellence” — eerily reminiscent of the accusation levelled at Microsoft by Steve Jobs more than a century later; as well as the lesson recently learned by digital artist Ben Moran when accused of posting AI-generated art to a subreddit.
Whether it’s digital artwork, ridesharing apps, social media posts or ancient textiles, one thing appears evident: the mass market doesn’t value originality as much as creatives would like.
Finally, what does ChatGPT itself say about the issue..?
It wouldn’t be an article on generative AI without asking ChatGPT a question, so here’s its response to “explain the intellectual property implications of generative artificial intelligence for artists and other creative workers”:
Generative Artificial Intelligence (AI) has the potential to raise various intellectual property (IP) issues for artists and other creative workers. One issue is that works created by generative AI may be considered “joint works” under copyright law, with the AI being considered a co-author. This could make it difficult for artists to exclusively own and control the rights to works created with generative AI. Additionally, it may be difficult to determine whether an artwork is original enough to qualify for copyright protection if it was generated by AI.
Another issue could be that AI-generated works may infringe on existing copyrights by creating works that are similar to existing ones. It could be difficult for copyright holders to enforce their rights against AI systems and their creators.
Furthermore, AI-generated works may also raise trademark issues if they depict or use protected logos, characters or other elements.
Overall, the use of generative AI in the creative industries is still a relatively new and evolving area, and there may be a need for legal frameworks to be developed to address the specific IP issues that it raises.
Subscribe to our monthly newsletter to stay up to date on research and upcoming events ✉️
