OpenAI is preparing to release a new artificial intelligence model called ‘Strawberry’, which promises to bring significant advancements in AI reasoning. According to a source familiar with the project, this model is designed to handle multi-step tasks, solving complex math and coding problems that current AI models sometimes struggle with.
What Sets Strawberry Apart
Unlike earlier versions of OpenAI’s ChatGPT, Strawberry’s unique feature is its ability to reason like humans. For instance, the AI can pause for a few seconds, analyzing related prompts behind the scenes before delivering what it determines is the most accurate response. This technique, known as “chain of thought” prompting, allows the model to perform tasks more accurately than previous AI systems.
Tackling Complex Problems
The Strawberry model can tackle challenges that typically stump AI models. For example, when asked if 9.11 is greater than 9.9, the model accurately identified 9.9 as the larger number. This enhancement represents a significant improvement in AI’s ability to perform multi-step reasoning tasks, which are often difficult for today’s systems.
Ongoing Development
OpenAI has been focusing on improving the AI’s reasoning abilities for some time. In May 2023, the company shared research on how it rewards AI for each correct step in solving math problems. The aim is to enhance AI’s multi-step problem-solving skills, rather than just rewarding the correct final answer.
Competition Heats Up
As OpenAI prepares to release Strawberry, it faces stiff competition from other tech giants like Anthropic and Google, both of whom are also developing AI models with similar reasoning capabilities. OpenAI’s model is expected to set a new benchmark in the field of artificial intelligence.
While no official release date has been confirmed, sources suggest that the Strawberry AI model could be made available to select users as soon as this week.