It's interesting that in last week's edition of The Bottom Line, Jane Friedman began with a note that was essentially “sorry there's so much on AI but it really is the big news right now.” Which is how I feel a lot of the time. I endeavor to keep AI news to one of the three weekly slots, but it's not always possible. Especially when, as today, it's a stuffed first slot and I have no idea what will break later in the week.
ALLi News Editor Dan Holloway
I'll start with the latest from the Anthropic settlement, because I know that will affect many of you in a very direct way. The time allowed for filing information to support your claim has been extended from thirty to sixty days.
OpenAI Knew, Documents Suggest
Meanwhile, there's a development in another case, this time involving Microsoft and OpenAI. In a turn that's reminiscent of the infamous internal memos from Meta, newly unsealed documents suggest that OpenAI knew they were training their GPT on pirated works and that creators would not be happy. Having just finished reading Karen Hao's fascinating Empire of AI, a gripping account of the leading AI companies, I feel like this isn't news. But having the actual memos with words such as (courtesy of Publishers Weekly, which notes the quotation comes from the complainant motion on behalf of the class) “We trained GPT-3 on pirated stuff! No sharing that!” is quite the thing. These quotations feel quite raw to read as a creator, but that does not of course mean that courts will find in favor of the class action the Authors Guild is bringing.
SCBWI Tightens AI Policy
Meanwhile, the question of AI use by writers continues to create waves. SCBWI (the Society of Children's Book Writers and Illustrators) has added a significant chunk to its policy on generative AI, with six additional rules that require authors and illustrators not to use AI in creating their work and to make a declaration of that lack of use. Award entrants will also be asked to keep records that could be used to demonstrate the humanness of the creative process, in a move that seems to be becoming more widespread as a protection against allegations against competition winners and organizers.
Which brings us finally to Pangram, the tool used by many armchair AI detectives. Most articles written about allegations of AI use rightly point out that AI detection software is liable to making mistakes. But a new article by Vauhini Vara suggests Pangram may be more accurate than people think. The piece contains detailed interviews with Pangram seniors, who obviously have a reason to support their product. But it also notes that Tuhin Chakrabarty, whose study of self-published titles for evidence of AI use sparked the recent round of debate, tested fifty self-published works from before the release of ChatGPT and Pangram came back with no false positives.
Thoughts or further questions on this post or any self-publishing issue?
If you’re an ALLi member, head over to the SelfPubConnect forum for support from our experienced community of indie authors, advisors, and our own ALLi team. Simply create an account (if you haven’t already) to request to join the forum and get going.
Non-members looking for more information can search our extensive archive of blog posts and podcast episodes packed with tips and advice at ALLi's Self-Publishing Advice Center.
