OpenAI is developing a “reflective” AI – Reuters writes about the Strawberry project

The project to develop a “superhuman artificial intelligence” was previously called Q* (Q-Star), the agency recalls. And according to sources, the Open AI management believes that thanks to it, it has moved on to the second of five steps on the path to creating “artificial general intelligence”.

According to Reuters analysts, work on this particular project and team disagreements over the issue may have been the reason for the firing (and subsequent reinstatement) of company founder Sam Altman last year.

The problem with current neural networks and AI robots is that they provide unreliable information in cases where more than a simple internet search is required. They cannot solve logical problems even at the level of playing tic-tac-toe.

The new form of AI will be able to “mirror” and surpass this threshold of capabilities, thus moving closer to more useful functions for humanity. For example, experts believe that thinking AI will be able to solve multi-level problems and even make scientific discoveries.

“We want our AI models to see and understand the world in the same way we do. Constant research into new AI capabilities is a common practice in the industry and everyone is convinced that over time these systems will improve in their logic,” commented an anonymous source from the agency on the development of Strawberry.

On the other hand, Reuters points out that the project is kept a top secret not only from the public, but also within OpenAI.