Episode Details

What if AI safety benchmarks are flashing a warning that the jobpocalypse is already starting? In this condensed summary of Doom Debates, host Liron Shapira talks with Center for AI Safety researchers Adam Khoja and Richard Ren about the pace of artificial intelligence, remote work automation, AI alignment, and catastrophe risk. In the full episode, they unpack how benchmarks like MMLU gave way to tougher tests such as Humanity’s Last Exam, the Remote Labor Index, and the MASK benchmark for honesty and deception. You’ll hear why they think model capability is improving faster than many safety measures, how AI could reshape jobs, robotics, and national security, and why U.S.-China coordination may be necessary to slow dangerous competition. This shorter recap preserves the key arguments and takeaways from the original episode so you can get the core ideas in minutes instead of the full listen. Listen now to get the key ideas in minutes.