OpenAI will release its new AI model ‘Strawberry’ earlier, possibly within two weeks. The super-intelligent AI model wasn’t supposed to be released until this autumn. Still, in the race to get to ‘general artificial intelligence’ as quickly as possible and stay ahead of the competition, CEO Sam Altman’s company has thrown the release of this new model into sixth gear.
This is according to anonymous sources to The Information. Strawberry would distinguish itself from OpenAI’s existing AI models and its competitors through highly advanced reasoning capabilities. The model is supposedly able to handle complex tasks such as programming, solving mathematical problems and even devising business strategies.
Smart but slow
The disadvantage of such thinking power is that it takes time. According to the anonymous sources, Strawberry takes up to 20 seconds to generate answers, as it considers accuracy more important than speed. This means this model will really work differently from the current ChatGPT versions. With those models, near-instant answers are the norm.
Speaking of ChatGPT, Strawberry is expected to be integrated as a standalone option into this platform. Using it will presumably carry a different price tag than the other LLMs. Also, users might get more or less message volumes depending on their subscription or faster responses when paying for a premium tier.
Short of memory
Another limitation of the model is that, for now at least, it can only handle text queries, unlike the multimodal capabilities of GPT-4o. There would also be delays even when answering simple queries. In addition, the model seems to have limited memory, making it less likely to remember previous conversations.
According to sources, Strawberry, which formerly went by the name Q*, played a role in last year’s crisis at OpenAI, which forced CEO Sam Altman to step down temporarily. The coup attempt was initially supported by co-founder Ilya Sutskever, who has clashed with Altman many times over OpenAI’s strategy.
AI geniuses neck-and-neck
Sutskever later said to regret his actions. His sudden turnaround left him as the only founder on the board after Altman’s resignation was withdrawn and it turned out to be the very board of directors that had to step down.
Some time later, Sutskever left OpenAI anyway, seemingly of his own free will, to found his own AI startup Safe Superintelligence Inc. Perhaps speeding up the release of Strawberry is a response to ‘genius’ Sutskever now being hot on the heels of OpenAI with building ‘artificial general intelligence,’ i.e. AI that approximates human brainpower. The fact that OpenAI has pressing money problems undoubtedly also plays a role.
Read also: OpenAI prepares ‘Strawberry’ model under high stress