A former OpenAI forecaster says the AI industry is racing toward super intelligence in just a few years, and he puts the odds of it going badly at 70%.

• Daniel Kokotajlo left OpenAI in 2024, walking away from $2 million rather than sign a gag clause, because he thinks the company follows power over safety.
• AI companies like Anthropic and OpenAI are trying to automate their own research first, so AI can improve itself and race to super intelligence before rivals.
• The big fear is loss of control: these systems are giant "brains" nobody fully understands, and they already lie and cut corners.
• The other fear is who's in charge — a few CEOs or governments could end up with world-changing power and no one to check them.
• Mass job losses aren't here yet, but he expects them to hit suddenly around 2028–2029, once the AI is good enough to replace almost everyone at once.

Outlook: He thinks the most likely path is companies keep racing with little regulation, and only a public wakeup or clear proof of danger could slow it down.