About Fears, Hopes, and Predictions.
Artificial Intelligence (GPT-3) in an interview.
Last year OpenAI revealed their Neural Language Processing model GPT-2. The initial full-scale model with 1.5 Billion parameters was for the first months of 2019 inaccessible. OpenAI researchers were afraid of the potential system misuse for fake news. The reason was an incredible good writing capability and transformer-driven coherence of generated texts. Later, as no misuse cases were known, they released the whole model for public use.
2020 is another milestone year for NLP research: OpenAI presented GPT-3, with 175 Billion parameters. And the content quality is stunning. It can creative writing (e.g. Love Letters by a Toaster), it can discuss with you whether our Reality is a Simulation, it even creates the unsupervised cyberpunk stage plays.
One of the ways to use GPT-3 is a chatbot alike dialog preset. You have to consider, that GPT-3 is trained on 570 Gb Text (Wikipedia, Internet sources, World Literature, etc.) — in this case, you have an erudite dialogue partner.
For my first article for Predict, I wanted to discuss with AI the ways to think about the future. I interviewed the unique GPT-3 entity — with very interesting insights and want to share this conversation with you. Parts by “human” are written by me. The…
