Ylanite Koppens / Pexels

Even though it is reportedly set to spend a whopping $7 billion on training and inference costs, with an overall $5 billion shortfall, OpenAI is resolutely striving to build the very first Artificial General Intelligence (AGI) in the world. Project Strawberry is the company’s next crucial step towards this lofty goal.

• What is Project Strawberry?
• What can Strawberry do?
• Why is it called that?
• When will Strawberry be released?
• Wait, what about its ability to autonomously navigate the internet? Isn’t that how we ended up with Ultron?

What is Project Strawberry?

Project Strawberry is OpenAI’s latest and potentially most remarkable large language model, expected to far surpass the capabilities of current state-of-the-art systems with its “human-like reasoning skills” when it is finally unveiled. It might even power the next generation of GPTs.

What can Strawberry do?

Project Strawberry is said to be a true reasoning powerhouse. It will have the ability to solve math problems it has never encountered before and act as a high-level agent, creating sophisticated marketing strategies and autonomously tackling complex word puzzles like those found in the NYT’s Connections. According to internal documents seen by Reuters in July, it can even “navigate the internet autonomously” to conduct “deep research.”

The Reuters report also notes that Strawberry’s architecture is similar to the Self-Taught Reasoner (STaR) technique. Developed at Stanford in 2022, STaR enables a model to generate its own training data on which to fine-tune itself, continuously enhancing its capabilities over time.

Why is it called that?

We don’t have the exact reason for the name “Strawberry” as OpenAI hasn’t publicly disclosed it. It’s a code name chosen for internal reference and to maintain secrecy during the development process.

i love summer in the garden pic.twitter.com/Ter5Z5nFMc

— Sam Altman (@sama) August 7, 2024

However, Strawberry didn’t start out as Strawberry. It used to be known as Q* (pronounced Q-Star), and during a particular time, Q* was the linchpin of CEO Sam Altman’s brief ouster last November.

OpenAI researchers specifically mentioned Q* in their letter to the company’s board, highlighting the potential risks posed by unregulated advanced AI. After Sam Altman regained control of OpenAI, Q* was reportedly renamed Strawberry in July 2024.

When will Strawberry be released?

According to The Information, Strawberry could potentially be released sometime in the fall. This is just a report, of course, so we can’t be certain for sure.

Wait, what about its ability to autonomously navigate the internet? Isn’t that how we got Ultron?

No, we got Ultron thanks to the comic book legends Roy Thomas and John Buscema, but the fears of an advanced AI growing out of human control are no longer entirely unfounded fantasies. Multiple sets of former OpenAI researchers have spoken out against the company’s efforts to develop an AGI, citing the lack of proper developmental safeguards.

Nevertheless, these complaints haven’t significantly slowed OpenAI’s drive to develop an AGI within the next decade. The company recently released a five-tier scale for measuring the capabilities of the AI systems it builds towards this goal and is currently seeking additional funding (yes, other than Microsoft) in a new round led by Thrive Capitol, with the company valued at over $100 billion.