OpenAI Prepares Big Upgrades For ChatGPT And GPT-4 With AI Codenamed 'Strawberry'

OpenAI will reportedly launch a new AI dubbed Strawberry (photo: x @Whizz_ai)

JAKARTA - OpenAI will reportedly launch a new AI dubbed Strawberry as part of a chatbot, most likely for ChatGPT, this fall. This AI has more advanced mathematical and programming problem solving capabilities than any other big language model (LLM). This makes it a great potential for performance improvement.

Strawberry, formerly known as Q Star, was first developed by former OpenAI chief scientist Ilya Sutskever and later upgraded by researchers Jakub Pachocki and Szymon Sidor. His most prominent ability is advanced mathematical reasoning, allowing this AI to solve unprecedented mathematical problems, an achievement that is often a challenge for today's chatbots.

The smaller and simpler version of Strawberry will be part of OpenAI chatbots, with the aim of maintaining larger model performance while pressing costs and increasing efficiency. While it is not yet clear whether this version will provide a significant performance boost on ChatGPT or GPT-4, its presence is expected to improve AI capabilities in more specific areas such as mathematics and programming.

Large version of Strawberry is used by OpenAI to produce high-quality training data, known as "synthetic" data. Using this data, OpenAI can reduce dependence on text and images from the internet, as well as minimize frequent errors in AI, such as hallucinations. Strawberry also plays an important role in the successor training of GPT-4, codenamed Orion.

OpenAI has presented Strawberry technology to US national security officials as part of efforts to be more transparent towards policymakers concerned about AI power. The move comes after several security leaders leave the company, raising the need to reassure the public and the government of OpenAI's responsibility in AI development.

While there is no exact date when the small version of Strawberry will launch this fall, this development is highly anticipated by the industry, wondering if OpenAI can maintain its lead in the AI race against its competitors.