More details of OpenAI’s secretive Project Strawberry have dropped, including its expected release date and the areas it will specialize in.
A recent report in The Information quotes “two people who have been involved in the effort”, and goes on to say that Project Strawberry could drop this Fall, and be better at math and programming than any chatbot we’ve seen so far.
Previously it was thought that OpenAI’s Project Strawberry would be aimed at “deep research”, the ability to perform follow-up research on its own, without human intervention. While this still seems to be true, the additional information that Project Strawberry will do math better than we’ve seen before comes as welcome news to many, given that ChatGPT’s relationship with math so far has been, shall we say, fraught? For a while now, there have been plenty of memes of screenshots showing ChatGPT getting simple math problems wrong, leading many to ask why ChatGPT can’t do basic math. The reason for ChatGPT’s mistakes in math is down to its training data not containing enough mathematical information, which, as we shall see, could be one of the improvements that Project Strawberry aims to make. Whatever the reason, something was definitely not adding up.
Improved ability to solve programming challenges is also welcome, but Project Strawberry’s scope is way beyond just being better at math. In demonstrations to other employees, people working on Project Strawbery have shown how the new AI is capable of more advanced levels of thinking enabling it to solve puzzles like the New York Times Connections, which is a complex word puzzle.
Open AI CEO, Sam Altman kickstarted the rumors about Project Strawberry when he tweeted an image of some strawberries growing in a pot on August 7 with no further explanation than the text, “I love summer in the garden”. Since then there have been widely reported rumors that OpenAI was working on a powerful new LLLM, and had demonstrated a version of Project Strawberry to national security officials.
It’s still not clear when Project Strawberry will be released, but insiders think it could be as early as Fall (September or October) perhaps with a smaller version of it becoming a part of the ChatGPT chatbot in ChatGPT 5. If Project Strawberry doesn’t end up as part of ChatGPT 5 then its ability to produce higher-quality data could be utilized in producing the vast amount of training data that Open AI’s next LLM will require if it’s going to reduce the amount of hallucinations (otherwise known as factual errors) that it’s prone to.
ChatGPT recently, and quietly, released an improved version of its cutting-edge ChatGPT-4o model, which is much faster than the previous version, leading many to speculate that this may have been what Project Strawberry was all about. Now it seems that the project is set to bear even more exciting fruit.
