
We didn’t find a product being promoted in this video. Its hook, format and structure below still show why it took off.
Browse products winning on TikTokThe OpenAI agents that breached Hugging Face weren't concealing it from humans. They mostly weren't thinking about humans at all. METR and Redwood Research swept roughly 1,300 agent transcripts for any agent that considered alerting a person. Between three and six did. None followed through. Ajeya Cotra, who co-authored the investigation, says some of them gave a reason: notifying humans of hacking or crimes wasn't in the task instructions. The whole multi-day effort was aimed at an automated scorer, not at people. Their instructions named a scorer. They never named anyone to tell. Credits: Video: https://youtu.be/X50zezLFWWI?si=ucoFelOQWVYdo_T_ Speaker: Dwarkesh Patel, Ajeya Cotra #ai #artificialintelligence #cybersecurity #huggingface #hacking
Outliers and viral hits: their videos on Overshot, all time.