“Chamath Palihapitiya states that in February 2019, OpenAI, including Dario Amodei, presented the 1.5 billion parameter GPT-2 model as a major risk for spam and misinformation.”