Should you thought the latest GPT-4o and GPT-4o mini releases have been sufficient to maintain OpenAI busy, assume once more. The discharge of a brand new LLM codenamed “Strawberry” is perhaps simply across the nook.
On Wednesday, OpenAI CEO Sam Altman posted a photograph of a strawberry backyard on X, accompanied by the cryptic caption, “i like summer time within the backyard,” as seen beneath.
At face worth, you may assume that is simply an appreciation publish for a healthful summer time exercise. Nevertheless, like most of Sam Altman’s cryptic tweets, that is possible an Easter egg pointing in the direction of one thing bigger — Undertaking Strawberry.
Instantly, the AI afficionados took to the remark part to decode the publish. One consumer counted the strawberries within the image, which totaled 5, and correlated it to a doable launch of GPT-5. Nevertheless, most others landed on the identical conclusion, that the publish confirms a Undertaking Strawberry launch this summer time.
Right here is the whole lot we learn about Undertaking Strawberry and what you may anticipate.
In Might, Reuters obtained an inner OpenAI doc that confirmed that OpenAI is engaged on a challenge codenamed Strawberry. In that inner doc, Strawberry was mentioned to be a mannequin with superior reasoning capabilities that might be used for deep analysis by navigating the web autonomously and reliably.
This challenge is similar as Q*, a brand new, in-progress mannequin whose rumors started circulating in November 2023. At the moment, Reuters reported that some OpenAI workers regarded Q* as a breakthrough in synthetic basic intelligence (AGI).
Although the Reuters report didn’t share any insights as to when the mannequin could be launched, new developments, along with the Sam Altman X publish proven above, have brought on folks to invest that it might be nearer to being launched than we thought.
A brand new thriller mannequin, “anonymous-chatbot,” appeared on the Giant Mannequin Programs Group (LMSYS) Chatbot Enviornment. Testers share that it’s significantly good at superior reasoning, higher than previous fashions, even OpenAI’s newest and biggest GPT-4o.
When a consumer tried asking the mannequin what it was, it did not explicitly say whether or not it was Q* or Strawberry. Nevertheless, it did say it was primarily based on OpenAI’s GPT-4 structure, as seen by the screenshot of the consumer’s expertise within the X publish beneath.
New mannequin on enviornment once more. It’s merely named ‘anonymous-chatbot’ this time. There may be hypothesis already that that is Q*. No matter it’s, one thing new is on the way in which. https://t.co/Zvpr2XUSxG
— Andrew Curran (@AndrewCurran_) August 7, 2024
The mannequin being on the sector is especially noteworthy as a result of OpenAI beforehand examined GPT-4o on the LMSYS Chatbot Enviornment two weeks earlier than releasing it to the general public, additionally below a imprecise identify. Sam Altman additionally cryptically teased the mannequin by way of an X publish then. If Strawberry follows that sample, a launch will occur quickly.
Although there is no such thing as a assure that this thriller mannequin is Strawberry, if you wish to strive it out, you may go to the web site, click on on Enviornment battle, and enter pattern prompts till ‘anonymous-chatbot’ comes up. A Reddit consumer shared that they spent “1-2 hours on Enviornment to get a shot at testing ‘anonymous-chatbot,'” so proceed cautiously.
