Support This Website! Shop Here!

Friday, February 03, 2023

Do AI Essays Violate Copyright?

 I'm not a lawyer, I do not play a lawyer on television, however, there is reason to believe that any essay generated by an AI is not subject to copyright prosecution.

According to the copyright office

Copyright law does not protect ideas, methods, or systems. Copyright protection is therefore not available for ideas or procedures for doing, making, or building things; scientific or technical methods or discoveries; business operations or procedures; mathematical principles; formulas or algorithms; or any other concept, process, or method of operation.

Section 102 of the Copyright Act (title 17 of the U.S. Code) clearly expresses this principle: “In no case does copyright protection for an original work of authorship extend to any idea, procedure, process, system, method of operation, concept, principle, or discovery, regardless of the form in which it is described, explained, illustrated, or embodied in such work.” Inventions are subject matter for patents, not copyrights. 

An AI typically assigns every item (in ChatGPT's example, that would be each word) in its training data set a numeric value. Once it has this string of numbers, the program then uses an algorithm to generate a new string of numbers which are converted back into words. Technically, the conversion and algorithmic operations upon the training data set arguably means the words of the resulting AI-generated essay are not copied, even if the resulting text is identical to one or more of the sections of the training data set. 

Furthermore, since this algorithm is a "business operation or procedure" that uses "mathematical... formulas or algorithms" in a "process, or method of operation", the conversion process is likewise not something that can be copyrighted. The specific code that expresses the algorithm can be copyrighted, but the conversion algorithm itself cannot be copyrighted.

If AI-generated essays turn out not to be subject to copyright infringement due to its algorithmic foundation, that raises an additional question. Why should the algorithmic operation of a machine be privileged over the algorithmic operation of the wet-ware which operates within an individual's brain? We may not fully understand how the brain's algorithms work, we certainly cannot replicate those algorithms, but we can obtain very similar response outputs by using them ourselves. 

So, do copyright and plagiarism rules seem reasonable in the 21st century? As knowledge becomes more of a computer-generated resource, it is hard to see how these ideas can continue to be useful within the culture.

No comments: