Paradoxically, it seems that the most important benefit of language models is in detecting their own fake text.

Superb! As a lifelong-lover of both Jurassic Park and the Santa Fe Institute, it oddly never occurred to me to explicitly connect Malcolm’s prescient rants with the surprisingly-senseless AI genies we’re releasing from the bottle.

But in 2017 I addressed the idea of training GANs to identify each other’s forgeries in the sci-fi novelette “An Oral History of the End of ‘Reality’.” Might be amusing to you, Dr. Mitchell… ;)