Several AI projects like ChatGPT, Llama, and Gemini have gained attention with their text-generating abilities. So far these models have gained recognition for their efficacy. However, these models have posed a hurdle in producing deep reasoning-based answers to the queries. Recently, it has been reported that Open AI has been working on a new AI model called ‘Strawberry’, which is yet to be disclosed officially by their team.
It was noticed that in July, Open AI was working on a project that could offer a distinct approach to handling data. This secret model by the AI Company showcases superior problem-solving abilities and exceptional reasoning skills. The project, earlier known as Project Q* (Q-Star) seems to be developed to answer all sorts of logical fallacies and common sense problems.
According to the sources, a new project model has been observed answering tricky Maths datasets and science questions. It is yet to be said if the model delivering such results is the Strawberry project or not. A few weeks back, the Open AI team conducted an internal meeting discussing a project that exhibits human-like reasoning ability. Although there has been no confirmation, it was expected the meeting talked about the Strawberry project.
It is expected that Open AI will integrate the Strawberry model with ChatGPT, which could make it convenient to use and accessible to everyone. The advent of the model could be a significant contribution to the AI industry, following which the model could be utilized in solving complex problems and enabling human-like thought processes. With efficiency, the Strawberry model can deal with logic-based text input and educate itself on the data, which could further improve Open AI’s new model development processes.
Earlier this year, Open AI CEO Sam Altman said the focus for the new AI models will be towards enforcing reasoning ability. The major capabilities of the model involve long-horizon tasks, where the model will plan future tasks and perform a series of actions to bring those plans into action. The model will be fed by ‘deep-research’ datasets, which will guide it in performing complex functions that existing AI models cannot perform. Strawberry’s model can break the barriers and push itself beyond the existing limits. The Open AI representatives also discussed the varied functions that can be done with assistance from the Strawberry model.
Several other tech giants like Meta and Google are also aiming to harness the potential of AI to develop reasoning-based models. Experts are yet to assure the effectiveness of such models, and if they will be able to predict and plan data just like humans.
There have been a few activities noticed where Open AI hinted to the developer community about the release of a more potent and effective AI model that can perform reasoning abilities. The Strawberry has a “post-training” generative model that can adapt and produce results according to the situation. Unlike other chat models, Strawberry does not only use the existing fed data but trains itself to plan and navigate the responses.
The AI giant is training its model to conduct a web search voluntarily by using computer-using agents. This can allow the model to decide its next step and generate responses without using any existing model datasets.
There has been a lot of haze around the viability of the project Strawberry. Various instances suggest Q* The demo model could be further flipped into the Strawberry project. The Open AI team signaled their involvement in the development of reasoning-based AI models. However, they are yet to make any announcement. The project development is kept secretive and most of the information is based on sources and reports by media agencies.