@damonmatthews.bsky.social 🧵10/9

While Rahal’s conclusion is optimistic, it may be necessary to hold this optimism to motivate the research and meet these ethical conditions. The AI era poses an ethical tightrope in which we will be balancing optimism and pragmatism for years to come. 🧵9/9

I have many concerns about the use and abuse of generative AI. I’ve read countless AI generated “undergraduate” papers and have concern for the loss of thinking skills, yet I see the value in making information more accessible and empathise with the need to get answers quickly in a fast world. 🧵8/9

“Condition 1: AI that uses public texts should be freely available to the public; Condition 2: AI must not misinform and ought to educate its users; Condition 3: AI must anonymize data to protect the public’s safety.” Okay tech world, you have a LOFTY research agenda to get to! 🧵7/9

Rahal lands on the conclusion that AI does have a right to learn, though it is not absolute. In furthering the Kantian analysis, he poses three conditions for the public and the developers behind the algorithms. 🧵6/9

Authors write publicly to be seen publicly, therein lies their autonomy and principle. In seeing authors as autonomous creatures and as ends in themselves- deeming publicly available text as being little but food for the AI machine undermines the intent and dignity of the author. 🧵5/9

Legally speaking, copyright has always been complicated. The Berne Convention and the Fair Use Convention are the places to look, the latter of which was used by both Microsoft and Google as it allows copying without permission as long as it adds value or transforms it. 🧵4/9

A defense heard in courts and dinner parties alike: Children, too, collect and store information to learn and grow. Rahal argues the distinct difference between a child and a generative AI training model is that the AI creates a digital copy of the text, as opposed to the child’s mental image. 🧵3/9

How a machine “learns” is by feeding on information…a lot of it. Not only does it know what the users feed it, it scans all publicly available text on the internet. News articles, discussion forums and comment sections, social media posts, etc. 🧵2/9