The program is able to generate quality texts using artificial intelligence so that its authors prefer not to publish it.
OpenAI is the research center in artificial intelligence (AI) supported by Elon Musk, Amazon and Microsoft and whose mandate is to distribute in the manner of the free software world its expertise to counterbalance private companies engaged in AI.
One of their new programs (GPT-2), an intelligent generator capable of producing a wide range of texts, news or essays is so realistic that a non-insider will not suspect its provenance.
In addition, this text generator can also translate and summarize documents.
Its authors at OpenAI went so far as to declare that their new software "achieves advanced performance on many criteria of language modeling".
And despite the original mandate of OpenAI, they decided not to publish or distribute the program to the public because of possible malicious applications if this technology was used by pirates or rogue states.
The deepfake of texts
They fear a proliferation of false news, online identity theft and automation of fake exchanges on networks.
As an example, the program was able to generate from a paragraph on a so-called flock of unicorns living in a distant and unexplored valley of the Andes a story of 300 words.
The authors fear that their GPT-2 program "has the same potential as the phenomenon of calls, images or videos deepfakes that abound on the networks."