OpenAI is reportedly working on more advanced AI models capable of reasoning and ‘deep research’

A brand new report from claims OpenAI is growing know-how to carry superior reasoning capabilities to its AI fashions below a secret challenge code-named “Strawberry.” Among the many challenge’s objectives is to allow the corporate’s AI fashions to autonomously scour the web with the intention to “plan forward” for extra complicated duties, in response to an inside doc seen by Reuters. The challenge beforehand glided by the identify of Q* (pronounced “Q star”), demos of which confirmed earlier this 12 months that it might reply “tough science and math questions,” Reuters stories, citing unnamed sources who witnessed the demonstrations.

At this stage, a lot stays unknown about Strawberry — together with how far alongside in improvement it’s, and whether or not it’s the identical system with “human-like reasoning” expertise that OpenAI reportedly demonstrated at an worker all-hands assembly earlier this week, per . However the capability for the corporate’s AI to conduct “deep analysis,” as is alleged to be the purpose of Strawberry, would mark an enormous leap ahead from what’s accessible at present.