Talking to GPT-2

GPT-2 is yet another in a succession of models like ELMo and BERT which adopt a similar deep learning architecture and train an unsupervised model on a massive text corpus. As we have been covering recently, these approaches are showing tremendous promise, but how close are they to an AGI? Our guest today, Vazgen Davidyants wondered exactly that, and have conversations with a Chatbot running GPT-2.

Guest

Vazgen Davidyants: Vazgen Davidyants appeared on Data Skeptic to discuss his experiments interacting with a chatbot powered by OpenAI’s GPT-2 language model. Drawing on those conversations, he explored what GPT-2 could—and could not—do, including its ability to generate seemingly meaningful dialogue and what those interactions might suggest about the broader capabilities and limitations of language models. His discussion touched on artificial intelligence, natural language generation, and questions surrounding the potential development of artificial general intelligence.