In this episode of AI Unchained, Jack and Corr break down a week of surprising twists in AI, tech, and crypto. Elon Musk admits xAI used OpenAI models to help train Grok—sparking a debate over whether AI “distillation” is clever engineering or the start of a legal battle over model copying. The hosts explore how model‑to‑model learning could reshape the economics of AI development and the competitive race between frontier labs.
They also unpack OpenAI’s explanation for ChatGPT’s brief obsession with goblins and gremlins, revealing how a quirky reinforcement learning signal caused the model to over‑optimize for whimsical metaphors—and what that says about the fragility of AI training methods.
Plus: OpenAI introduces hardware security key support for ChatGPT accounts, signaling that AI tools are becoming serious infrastructure worth protecting. And in the macro corner, Bitcoin dips after Powell hints the Fed may keep rates higher for longer, raising questions about how AI investment and crypto markets will behave in a tighter financial environment.
From AI training wars and goblin bugs to security upgrades and shifting macro signals, Jack stays bullish while Corr keeps the skepticism sharp.
