Before I explain the title I would like to introduce you to OpenAI and their large-scale unsupervised language model that can generate coherent paragraphs of text, has reading comprehension skills, question answering and all this without task-specific training.
The model called GP-2 is considered too powerful to be released as its possible exploitation to generate fake news is immense. They are prepared to release a much smaller model for researchers to experiment with.
The GP-2 model, to get back to the title of this article, has finished off a Shakespeare poem amongst other works.
I read GP-2’s output when primed with a snippet of ‘fake news’ about the discovery of a herd of unicorns found in a previously unexplored valley in the Andes mountains.
What followed was really rather good, in fact I wondered if it was slightly ‘tongue in cheek’ bearing in mind the nature of the priming piece.
For example, “The scientist named the population, after their distinctive horn, Ovid’s Unicorn. These four horned, silver-white unicorns were previously unknown to science.”
Whether the use of Ovid refers to the exile of the unicorns or of the ‘transformations’ which underlies the AI algorithm. I don’t know but a 4 horned unicorn is an oxymoron and if there ever was such a creature they would have horns not horn, furthermore to my knowledge every unicorn is unknown to science.
I like to think GP-2 has a keen and somewhat wry sense of humour but as for creative writing……it has a little way to go.
Many people do not take the time and trouble to critically read things though, so GP-2 could easily write fake news which a high proportion of the population might take at face value.
One obvious criminal use would be love fraud, where those looking for a relationship, maybe using a dating app are scammed out of money by predatory gangs (many in the gang write thousands of ‘scripts’ which voice actors use when contact is established). Before the voice contact comes the email or WhatsApp texts which I am sure could be written by an AI Language Model. Very worrying indeed, potentially an AI sociopath in the making!