’This Could Go Quite Wrong’ - Altman Testimony, GPT 5 Timeline, Self-Awareness, Drones and more

That was the blunt warning of Sam Altman at the start of his illuminating testimony in front of Congress. I picked out all the most interesting parts, from timelines for GPT 5, whether models are self-aware, whether we are training them to lie, what jobs are going and when, and far more. I’ll showcase some of the safety concerns raised at the testimony, and give background from Anthropic and Google Deepmind. The Senate asked a few interesting questions, and phrased them in strange ways, and I’ll cover that too. Palantir and their drone LLM will appear, as will the new Constitution from Anthropic for Claude Plus and other models, and Victoria Krakovna. I won’t cover much of Sam Altman having no equity in OpenAI, as that was mentioned in my $100 Trillion video!