ChatGPT maker OpenAI is reportedly working on a new model codenamed “Strawberry” to enhance the AI reasoning of its models and enable them to navigate the Internet independently.

Also read: OpenAI to join Washington tech lobbying group BSA

The project, whose details are closely guarded, is inspired by Stanford’s “Self-Taught Reasoner” and involves models creating their training data.

Strawberry to improve OpenAI models

A report by Reuters citing internal documents and people familiar with the matter indicates that Strawberry will significantly improve OpenAI’s AI models. The project was previously known as Q*, which Reuters reported was already seen as a breakthrough.

While finer details of the model have been kept under wraps even from staff, demos of Q* to some staff showed that the LLMs could answer some complex maths and science questions that are currently “beyond the reach of commercially available models.”

An internal document and source reveal OpenAI's Strawberry, formerly known as Q*, which aims to dramatically enhance its AI models' reasoning capabilities. Pushing the boundaries of AI! #OpenAI #Strawberry #AI #ArtificialIntelligence #TechNews pic.twitter.com/101IjnpLJI

— PUPUWEB Blog (@cheinyeanlim) July 12, 2024

According to the documents, OpenAI wants Strawberry to perform long-horizon tasks (LHT) which requires an AI model to “plan ahead and perform a series of actions over an extended period of time.” Specifically, OpenAI wants its generative AI models to not only answer users’ queries alone, but be able to autonomously search the internet and reliably perform what the company calls “deep search.”

“We want our AI models to see and understand the world more like we do,” said an OpenAI spokesperson who spoke with Reuters.

“Continuous research into AI capabilities is a common practice in the industry, with a shared belief that these systems will improve in reasoning over time,” added the spokesperson.

The company hopes Strawberry can achieve this through extensive “post-training” analysis of its existing AI models to generate more human-like responses. It is not yet clear when Strawberry will be made available to the public, as it is still work in progress.

Elon Musk shows confidence in OpenAI’s Strawberry

Meanwhile, the tech billionaire showed confidence in the project, and the future of AI. In a post on X platform, Musk juxtaposed the doom and gloom that AI would supposedly pose to humanity versus its bright prospects. Musk hinted at AI’s potential, far from being a threat to humanity.

“We were told AI doom would be a paper clip maximizer, but it’s actually going to be Strawberry Fields Forever,” he wrote.

Also read: Microsoft relinquishes OpenAI board seat amid antitrust scrutiny

According to Coingape, Musk’s remarks are a sign of confidence, while the phrase “strawberry fields forever” alludes to “ideas of a utopian outcome with AI development, potentially inspired by the Beatles song.” Musk’s remarks also attracted attention across the industry.