This New Theory on How AI 'Thinks' Could Impact Whether It Kills Us - Business Insider

Nick Cammarata, an independent researcher who formerly worked at OpenAI and said he left the company in 2023, published a report on Thursday, according to Business Insider. The article states that some in the AI world are describing the report as a bombshell. The report argues that AI neural networks are knowable so long as you look carefully. Business Insider published the article under the headline "This New Theory on How AI 'Thinks' Could Impact Whether It Kills Us" on businessinsider.com.

According to the article, the report concerns how AI models think. Cammarata's stated position is that neural networks can be understood through careful examination rather than treated as beyond inspection. The excerpt does not set out the specific method the report proposes beyond the claim that careful examination is sufficient to make such networks knowable. Business Insider notes that the subject matter is the process between what goes into a model and what comes out of it.

The article states that the finding could matter for safety researchers who are scrambling to understand what happens between model input and model output. The headline frames the theory in terms of whether AI could kill humans, connecting the question of how models think to existential risk. Cammarata is described as an independent researcher, with the article stating that he left OpenAI in 2023.