Geoffrey Irving Claims False Positives Shaped OpenAI Tradeoffs
UK AI Safety Institute chief scientist replies on differing predictions from safety reasons.
In a reply to @tomekkorbak, Geoffrey Irving states his claim that false positives were one of the reasons behind OpenAI safety decisions. He notes both reasons apply but make different predictions. One example he gives is that OpenAI will dial false positives down at the expense of false negatives. Irving is identified in the post as Chief Scientist at the UK AI Security Institute with prior roles at DeepMind, OpenAI, and Google Brain focused on alignment.
Combined views
199
1 post, first seen 6h ago
Geoffrey Irving Claims False Positives Shaped OpenAI Tradeoffs
UK AI Safety Institute chief scientist replies on differing predictions from safety reasons.
In a reply to @tomekkorbak, Geoffrey Irving states his claim that false positives were one of the reasons behind OpenAI safety decisions. He notes both reasons apply but make different predictions. One example he gives is that OpenAI will dial false positives down at the expense of false negatives. Irving is identified in the post as Chief Scientist at the UK AI Security Institute with prior roles at DeepMind, OpenAI, and Google Brain focused on alignment.
Reactions
1 postGI 
Geoffrey Irving@geoffreyirving@tomekkorbak Yes, and my claim is that they were one of the reasons. I think it is important to understand this: both reasons apply, but they make different predictions. For example, one thing my reason says is that OpenAI will dial false positives down at the expense of false…
- likes: 3
- replies: 0
- bookmarks: 0
- reposts: 0