There are growing rumors that OpenAI is set to unveil a new model this week, potentially marking the debut of the much-anticipated “Project Strawberry,” which is believed to focus on enhancing AI reasoning capabilities.
Following artificial intelligence discussions on social media lately has been reminiscent of childhood strawberry-picking memories, as numerous accounts—including that of OpenAI CEO Sam Altman—have been sharing images of strawberries.
Altman recently posted a photo of a strawberry plant with the caption, “I love summer in the garden,” showing four strawberries. This has led to speculation that a new reasoning-oriented version of GPT-4 could be in the works, possibly designed to operate alongside GPT-4o, which is geared towards creativity and interaction.
Initial leaks about Project Strawberry emerged in May, suggesting it represents a significant step towards artificial general intelligence (AGI), where AI could potentially mimic human-like thinking and behavior. However, it’s still unclear whether this will be an enhancement to the existing GPT-4 or an entirely new model.
What exactly is Project Strawberry?
While OpenAI has not officially disclosed details about Strawberry, leaked documents indicate it is akin to a system known as Q*. This involves training a model to make decisions by estimating the value of its actions.
An AI model with advanced reasoning capabilities could have transformative impacts, revolutionizing fields such as scientific research, healthcare, and finance, and driving technological innovation at an accelerated pace.
According to a May leak shared with Reuters, Strawberry is designed to address reasoning challenges that previous models have struggled with. For instance, it could enable ChatGPT to independently solve complex problems and verify its own outputs.
Even for everyday users, this model could bring significant improvements. It might offer better travel planning, more personalized health advice, and even help analyze spending habits to create a tailored budget for financial independence.
OpenAI has outlined five levels on the path to AGI, with the current level being chatbots—AI that can engage in natural conversations but with limited problem-solving ability. Strawberry is positioned as level 2, focused on reasoning.
Is this all just hype?
On X (formerly Twitter), a user named iruletheworldmo started sharing random images of strawberries, hinting at an impending “level two” breakthrough. They wrote, “welcome to level two. how do you feel? did I make you feel?” to which Altman responded, “amazing tbh.” This sparked a wave of Strawberry-related speculation, which was further fueled by Altman’s subsequent tweet featuring the strawberry plant.
Adding to the excitement is the fact that Google is holding its Made by Google event this week, and OpenAI has a history of making major announcements during Google’s events. This has only amplified the hype.
While it’s uncertain whether a new reasoning model is indeed on the horizon, and what form it might take if it is, there is a strong sense that we are approaching another significant leap in AI model capabilities.