I was planning to regurgitate some of the current discussions about AI, particularly with respect to copyright (this is, after all, a blog by and for authors), but I have just changed my title to Untrained Melody because of something I noticed this morning on Amazon.
One of the most important factors in buying something on Amazon is how many stars a book or other product has, and how many five-star reviews there are. Well, now that Amazon is using AI to sort reviews, one maybe cannot trust Amazon star ratings and reviews.
I am looking for a magic mat that will gently shock my feet and make foot and leg pain vanish temporarily. Walmart has something of the sort on clearance for less than $10, and it appears to have very poor reviews. So, of course I visited Amazon, willing to pay four or five times the Walmart price for a foot-shocker for authors and other sedentary people, providing that the Amazon product has good reviews.
Check this out, not literally, of course.
"Customers like the quality and sturdiness of the health personal care product. They mention that it is beautiful and well made.
AI-generated from the text of customer reviews"
The sentence about AI is in very fine print. (I put up a screen shot, but that was not permitted).
Intrigued for many reasons, I scrolled until I found the reviews. Here is the reviews link:
What should tires and a low-slung skirt, or a necklace, or a router table, or a sandbag have to do with a stimulating bit of electrified leather or plastic?
Maybe, authors, you should investigate what Amazon AI may have done to your reviews. Maybe, would-be-buyers, you should check if you are looking at a product where the reviews have been AI generated from the text of customer reviews.
I might have happened upon an isolated SNAFU.
Even so, the AI does not appear to have perfected the spelling, punctuation, grammar of the reviews, so what use is it?
The Mintz law firm's Insights Center has an excellent (if possibly over-punctuated) examination of the (Un)fair Use? Copyrighted Works as AI Training Data.
Authors Bruce D. Sokler, Alexander Hecht, Christian Tamotsu-fjeld, and Raj Gambhir discuss the dilemmas that have arisen around the world from the methods used by developers to train their tools and models.
As they write:
"Many have been amazed by the capacity of generative AI tools to
answer questions, crack jokes, and compose poetry."
And:
".. for training AI models,
researchers have “scraped” various data sources including internet
forums, book corpuses, and online code repositories."
Allegedly, it wasn't a contentious issue until AI became commericalized. Perhaps a use can be fair as long as it is for the benefit of all mankind, and is not exploited for the financial gain of a few.
Apparently, in Europe, there is a right for copyright holders to opt-out of having their work data mined, and if they do so, the AI trainers must seek permission. It would have been simpler if the AI trainers had stuck to works that are unquestionably in the public domain, but would they find Chaucer,
Aphra Behn, Shakespeare,
The Castle of Otranto, and the Bible a tad inconvenient for training purposes?
Meanwhile, there are lawsuits. Last month, the
New York Times sued leading generative AI companies
for “unlawful use of The Times’s .... content without permission to develop their
models and tools.”
The Authors Guild has organized a class-action lawsuit against OpenAI. There is an article about it on the Times site, here:
The purpose of the law suite is to take a stand against wholesale theft of the work of all authors, no matter the genre in which they write. The Authors Guild penned an open letter, which thousands of authors supported. You can use the search function to check whether or not you signed.
My name is on page 22 of 200 pages of supporters' names,
All the best,
Rowena Cherry
SPACE SNARK™
Posts
No comments:
Post a Comment