“‘Generative AI’ cannot generate anything at all without first being trained on massive troves of data it then recombines,” Joseph Gordon-Levitt writes at the Washington Post. “Who produces that training data? People do. And those people deserve residuals.”
In the sense that Gordon-Levitt — an excellent and well-known actor, writer, and director — uses the term, “residuals” are the payments actors, writers, and directors receive when productions they act in, write, or direct get re-used as television re-runs, streamed media, DVD releases, etc.
Even in that very specific context, residuals only go so far. They’re a specific benefit negotiated between unions (on behalf of their members) and entertainment production firms, not a general principle with obvious applicability to anything and everything one person (or entity, like an AI “large language model”) might happen to learn from another.
Most of the things we do every day are “residual” in the sense that they rely on the “residue” of a large body of knowledge developed over thousands of years by others.
Fortunately, we don’t have to mail a penny to the estate of whoever invented the wheel — circa 4500 BC — each time we jump in our cars or on our bicycles to go somewhere, or to the descendants of Shakespeare every time we suggest that someone doth protest too much (Hollywood screenwriters would pay through their noses if Shakespeare’s estate got residuals — almost any modern production works as “which of The Bard’s plays are they cribbing from?” fodder).
Nor, even stipulating to the idea of “intellectual property” as a valid concept, is the debt we owe those we learn from something that we traditionally pay “residuals” on.
How many teachers, friends, and loved ones helped make Joseph Gordon-Levitt the man — and the actor, writer, and director — he is? How many people did he learn from? Quite a few, I suspect … and because he seems like a good guy, I’d be surprised if he hasn’t thanked them for it as best he can, in both speech and in action. But 99.9% of them are likely not collecting, nor are they owed, residuals on Inception, Looper, and Snowden.
In terms of inherent financial obligation, what’s the difference between Joseph Gordon-Levitt learning his chops from reading great writing and watching great performances, or an AI learning its chops from reading great writing and watching great performances?
In my opinion, there’s neither any difference nor any payment due.
Thomas L. Knapp (Twitter: @thomaslknapp) is director and senior news analyst at the William Lloyd Garrison Center for Libertarian Advocacy Journalism (thegarrisoncenter.org). He lives and works in north central Florida.