Is OpenAIs Project Strawberry GPT-5 in the making?

OpenAI is working on a new project codenamed ‘strawberry’ that has been teased multiple times by CEO Sam Altman. A report by Reuters talked about the internal documents that revealed Project Strawberry will allow AI to perform detailed research on the web. Also Read: OpenAI Confirms Adult-Only ChatGPT With Custom Personalities And Erotic Conversations

One of the OpenAI’s spokespersons said that the latest systems in the pipeline are focused on improving reasoning. Rumors have it that Project Strawberry could be nothing but GTP 5 in development. Other leaks regarding the same suggest that the demo version was able to solve math and science-related problems, a new achievement for AI models. Also Read: UPI Meets ChatGPT: India Tests AI-Powered E-Commerce Payments Via OpenAI Partnership

Project Strawberry: GPT 5 in the making?

The Project Strawberry demo model clocked a 90% score in the Math dataset. Furthermore, OpenAI also hosted an internal demo, as reported by Bloomberg, in which it showcased a research project that performed human-like reasoning. Here are the five levels of AI progress presented by Open AI Also Read: OpenAI Bans Multiple ChatGPT Accounts: Know The Whole Reason

• Level 1: AI can chat like humans in natural language, also known as Conversational AI.
• Level 2: AI can reason the way humans do, dubbed Reasoning AI.
• Level 3: AI can perform actions an on autonomous basis, called as Autonomous AI.
• Level 4: AI starts innovating and improving with time, termed as Innovating AI.
• Level 5: AI can perform actions and get done with tasks for an entire organization, known as Organizational AI.

Keeping the current developments of OpenAI in mind, Project Strawberry marks its debut at level 2. Here, the AI system will have the capability of reasoning like humans. The model will also come with a post-training method that will help it in improving gradually. The technique used in the post-training of Project Strawberry is pretty similar to STaR or Self Taught Reasoner. STaR will let the model modify itself by learning from self-generated reasoning.

With all the development on hand, we can say with surety that OpenAI is setting up the stage to take the next big leap. As of now, it is not clear if the model will be known as GPT 5 or not. However, OpenAI CEO Altman has already mentioned previously that the successor of GPT 4 will most likely ditch the GPT 5 title.