As many of you might recall, Sam Altman, CEO of OpenAI, posted a photo of strawberries on social media earlier this month, sparking speculation about the imminent release of an AI model named "Strawberry."
Now, there's finally new information about "Strawberry"! According to The Information, OpenAI has demonstrated this secret project to U.S. national security officials, ensuring its development and use will be within safe and reasonable limits.
Insiders have revealed that this AI product is expected to debut this fall as part of a chatbot, potentially integrated into ChatGPT. Strawberry's standout feature is its ability to solve mathematical problems, which is rare among current chatbots. Moreover, it can handle programming issues and its applications extend far beyond these.
After gaining more "thinking" time, Strawberry can even address more subjective questions like product marketing strategies. OpenAI employees have showcased Strawberry's ability to solve the New York Times crossword puzzle, demonstrating its exceptional performance in language tasks.
OpenAI's efforts to launch Strawberry are part of its strategy to maintain its lead in the conversational AI field. This technology has profound implications for future products, such as multi-step task agents, which could open up new revenue streams for the company.
OpenAI's business growth is remarkable, with its master's degree and ChatGPT subscription revenue tripling year-over-year to $283 million per month. Despite this, the company's private valuation stands at a staggering $86 billion, with its prospects largely hinging on the successful launch of its new flagship large language model (LLM), codenamed Orion.
The chatbot version of Strawberry, a distilled edition, aims to maintain the same performance as the original model while being more user-friendly and cost-effective. Additionally, OpenAI uses a larger version of Strawberry to generate training data for Orion, known as "synthetic data," helping the company overcome limitations in obtaining high-quality data.
Using high-quality training data generated by Strawberry can help reduce errors produced by the model, known as "hallucinations." Imagine a model that can accurately answer logical puzzles without hallucinations—a significant advancement.
Key Points:
🍓 OpenAI's upcoming AI model focuses on solving complex math and programming problems.
🚀 OpenAI's master's degree and ChatGPT subscription revenue has tripled year-over-year, with a company valuation of $86 billion.
🔍 High-quality training data generated by Strawberry is expected to significantly reduce errors produced by AI models.