How much of the AI / Fair Use concern is about academic fraud?

Everyone will have seen concerns expressed about AI, on various grounds, in particular "alignment". But there's another line of attack: that ingesting text into the datasets used to train the AI models is a breach of copyright, and outside the "fair use" protections of the United States' Bill of Rights.

The thing is: I am old enough to remember campaigning precisely for wide exemptions to copyright to permit exactly this kind of re-use. Indeed, campaigning for the removal of *automatic* copyright on written works in the first place and returning to a system of copyright-by-registration, which would certainly make Gemini easier to operate legally.

(As an aside - Gemini effectively takes the assumptions about copyright that were abroad when the Web was being created in 1990, which was just after the change to automatic copyright in the United States.)

It seems that some of the campaigners have even completely switched sides from those days. The argument *used* to be that the so-called "moral rights" of the original creator were philosophically and economically unsustainable, and entailed mass privatisation of the intellectual commons. The right to re-use and re-mix material created by others was seen as an important protection for the general public, consumers, and even commercial creators. But where the re-use was not so transformative, and undermined the market for the original work, it was less likely to be permitted.

Those conditions obviously don't apply to LLMs that spit out text. It doesn't affect anyone's willingness to buy the works of Maya Angelou or Tom Wolfe or Aleksandr Solzhenitsyn or Brian Kernighan that their text has informed the weights used in ChatGPT.

But I think there's at least another thing going on: LLMs can automate the detection of various types of "fraud", certainly "malpractice", in academia: the fake citations, mistakes of logical inference, and manipulated data. And they can do so at scale.

Bill Ackman seemed to think so:

Bill Ackman suggests AI-powered plagiarism checks will cause ‘incredible embarrassment’ in academia

I wonder why it hasn't happened yet?

Gemlog index

Site index