Be part of our every day and weekly newsletters for the most recent updates and unique content material on industry-leading AI protection. Be taught Extra
Regardless of an government shuffle on the high of the corporate, OpenAI co-founder and CEO Sam Altman appears unbothered — a lot in order that he posted on a picture to his private account on the social community X immediately displaying a nice view of strawberries rising in planters in a backyard, captioned with “i love summer in the garden” in his typical all-lowercase textual content fashion.
Whereas the photograph by itself appears pretty innocuous and even just a little bland, it has Altman’s followers on X and others within the giant and energetic AI group on the positioning giddy with pleasure that Altman is definitely acknowledging one of many longest-standing rumors about OpenAI’s efforts to develop a brand new basis mannequin supposedly codenamed “Strawberry,” which can or might not be equal to the anticipated GPT-5 basis mannequin.
“What other clues do we need?” posted @kimmonismus on X, a number one AI rumor and knowledge dissemination account.
Bindu Reddy, CEO of open-source AI startup Abacus AI, additionally posted on X weighing in that it was a “reference to Project Strawberry.”
Additional including gasoline to the fireplace is the truth that a brand new “anonymous-chatbot” AI mannequin has been added to the favored open-source lmsys arena, the place quite a few AI mannequin supplier firms together with OpenAI itself typically check upcoming releases with customers previous to an announcement.
The brand new “anonymous-chatbot” when requested claims to be a part of OpenAI’s GPT-4 household of fashions, and follows comparable language to the “gpt2-chatbot” that appeared on lmsys earlier than the discharge of OpenAI’s GPT-4o multimodal mannequin in Might 2024.
Early customers of the brand new “anonymous-chatbot” report that it displays extra superior reasoning capabilities than GPT-4o and different rival frontier fashions, suggesting it might be extra highly effective than them.
Rumors of a massively highly effective new OpenAI mannequin kicked off with a report from Reuters in November 2023 describing a brand new AI mannequin referred to as “Q*” (pronounced “Q-Star,” in response to the report):
“Some at OpenAI imagine Q* (pronounced Q-Star) might be a breakthrough within the startup’s seek for what’s referred to as synthetic basic intelligence (AGI), one of many individuals informed Reuters. OpenAI defines AGI as autonomous methods that surpass people in most economically worthwhile duties.“
Reuters adopted by with an unique report in mid-July stating:
“ChatGPT maker OpenAI is working on a novel approach to its artificial intelligence models in a project code-named “Strawberry,” in response to an individual aware of the matter and inner documentation reviewed by Reuters.”
As for what particular new capabilities Strawberry could supply, Reuters reviews:
“The document describes a project that uses Strawberry models with the aim of enabling the company’s AI to not just generate answers to queries but to plan ahead enough to navigate the internet autonomously and reliably to perform what OpenAI terms “deep research,” in response to the supply.
As well as, it has been practically 17 months, or about 1.5 years, since OpenAI launched GPT-4, its final main new basis mannequin class.
And whereas it has launched new basis fashions GPT-4o and GPT-4o mini in that point — the primary educated to be natively multimodal moderately than natively textual content — the competitors from Anthropic and Google and Meta with equally or in some instances extra highly effective fashions, no less than in sure duties, has solely elevated markedly as effectively.
We’ve reached out to OpenAI for extra details about Altman’s publish on X and can replace once we hear again.