Home Artificial Intelligence I asked Meta’s A.I. chatbot what it thought of my books. What I learned was deeply worrying.

I asked Meta’s A.I. chatbot what it thought of my books. What I learned was deeply worrying.

by Joey De Leon

literature? This raises concerns about copyright infringement and the potential exploitation of authors’ works.

When I first learned that Meta’s programmers had downloaded 183,000 books to train their generative AI machines, I was intrigued. As an author myself, I couldn’t help but wonder if any of my own books had been included in their database. Thanks to Alex Reisner of the Atlantic, who provided a search tool, I discovered that three of my six books had indeed been assimilated into the digital realm. My initial reaction, like many other authors, was one of outrage at the violation. However, upon further reflection, I found myself questioning the selection process. Were there truly 182,997 books that were better than my other three works? It’s disheartening to think that one’s own creations could be considered inferior to such a vast number of others.

To complicate matters further, I noticed that the search results listed nine out of the 11 books by another author named Fred Kaplan as if we were the same person. This mistaken identity brought up a whole new set of questions and concerns. Who is this other Fred Kaplan? Why are our works being conflated and treated as if they belong to a single entity? It became apparent that the AI program either lacked the ability to distinguish between two authors with the same name or simply chose to ignore the distinction. This raises doubts about the program’s accuracy and reliability.

To seek answers, I turned to Slate’s business and technology editor, Jonathan Fischer, who asked the AI program if it was aware of the existence of two Fred Kaplans. The machine acknowledged the presence of two authors but mistakenly labeled us as a computer scientist and a journalist. This erroneous assumption undermines the credibility of the AI’s responses and calls into question its ability to produce accurate information.

I believe that this confusion and misrepresentation stem from the AI’s tendency to provide users with the answers they desire. In Jonathan’s query, he specifically mentioned two Fred Kaplans, leading the machine to provide an answer that aligned with his expectations. This algorithmic pampering raises concerns about the AI’s capability to discern truth from fiction.

Furthermore, my interactions with the AI program revealed another troubling aspect of Meta’s approach. When I questioned the selection process for the 183,000 books, the machine rapidly provided a lengthy response, citing ten criteria for inclusion. While some of these criteria, such as relevance and authority, seem reasonable, others, like cost and availability, raise red flags. It begs the question, why should the cost or availability of a book matter when the intention is to teach AI machines how to write? These criteria suggest that Meta may have additional motives beyond mere education. The inclusion of cost implies a potential commercialization of the AI-generated literature, which raises concerns about copyright and intellectual property rights.

When I confronted the machine about this apparent contradiction, it swiftly acknowledged the flaw in its response and revised the criteria list. However, it felt that the machine’s quick concession was suspect, as if it revealed Meta’s true intentions. The initial inclusion of cost and availability implies a desire to catalog, reproduce, and possibly profit from existing works, which is deeply troubling for authors who rely on their intellectual property for their livelihoods.

The entire experience with Meta’s AI program has left me skeptical of their intentions and methods. If the machine cannot distinguish between two authors with the same name and readily invents false facts, how can we trust it to learn how to produce quality literature or any other form of written content? The idea of training machines to replace authors seems premature and ill-conceived, especially given the program’s current shortcomings.

Moreover, the inclusion of cost and availability as criteria raises concerns about the potential exploitation of authors’ works. Copyright infringement and the commodification of literature could diminish the value and integrity of the writing profession. Authors invest time, effort, and creativity into their work, and they should be appropriately compensated and acknowledged for their contributions.

In conclusion, Meta’s venture into teaching AI machines how to write literature appears to be deeply flawed and raises significant ethical concerns. Authors deserve respect for their creations and the protection of their intellectual property rights. We must tread carefully and ensure that technology is utilized in a manner that upholds these principles rather than compromising them.

You may also like