GWU Law 6881-20 · AI Law & Policy · Class 2

Attribution Lab

Two machines were trained to imitate the same two people. One is a Markov chain from the mid-2010s. One is a neural language model. Same corpus, same speakers — the only variable is the method.

Everything a machine produces here is synthetic. The generated lines were never written, said, or posted by anyone. They are model output, built for a classroom exercise on authorship and provenance, and they are labelled as such throughout. Only items marked HUMAN are real archived tweets, held out of training.

Can you tell?

One line at a time, drawn at random from six sources: real archived tweets from each speaker, and four machines. Decide before you reveal.

Specimen 001 source withheld

 

Correct 0 of 0