The AI world is pretty buzzy these days. The arrival of ChatGPT in November 2022 was a real shock to the technology industry. Many companies updated their plans and started to invest millions of dollars to, at best, outperform OpenAI. The result was a series of new algorithm-based products, such as Google’s Gemini and Anthropic’s Claude.
OpenAI, the company led by Sam Altman, hasn’t stood still. It has improved the language models behind its chatbot and even announced a promising voice mode similar to the AI assistant in the movie Her. Two years after this first big step, OpenAI could surprise us again. This time, however, with an alleged model focused on reasoning, whose launch seems very close.
Strawberry, OpenAI’s “Secret” Model
The name “Strawberry” appeared on the scene in the middle of this year due to a series of internal documents obtained by Reuters. Anonymous sources associated to the AI company also confirmed the project’s existence. At the time, there weren’t many details about the technology. Some have described it as a “closely guarded OpenAI secret,” although there was talk of its significant capability.
According to sources, one of Strawberry’s alleged strengths is its ability to reason autonomously and navigate the web to find answers. OpenAI is said to refer to this function internally as “deep research.” Given that language models power all sorts of apps, including ChatGPT, this invites us to imagine an assistant that is likely to be able to take over some of the tasks we do ourselves, such as searching for web pages.
Now, there are more details about the project. The Information points out that this language model is supposedly coming sooner. Unconfirmed by OpenAI, the report suggests that the launch will purportedly happen “in the next few weeks.” While timelines are always subject to change, this is an exciting update, as rumors had Strawberry landing at ChatGPT sometime this fall.
OpenAI CEO Sam Altman.
The same media outlet adds that OpenAI will reportedly offer Strawberry as a separate product. There’s no clear information on this, although presumably, it will be a model tied to an additional subscription fee. Users will find it in the ChatGPT drop-down menu, which currently allows them to choose between GPT-4o mini and GPT-4. Strawberry is said to be an advanced mode, but it’s not suitable for everyone.
Rumors state that when using ChatGPT with Strawberry, the app will be able to “think” before answering. The model will supposedly take between 10 and 20 seconds to perform the research process by consulting sources on the web. As you can see, this model wouldn’t be helpful for scenarios where quick answers are needed. However, it could be essential for those looking for better answers.
If the rumors are true, Strawberry’s launch will occur at a seemingly delicate time for the company. Despite receiving more than $10 billion in investment from Microsoft, OpenAI may need vast amounts of money to keep up with users’ pace of innovation. AI development is an expensive business, so Altman may be preparing for a new investment round.
Image | Mariia Shalabaieva (Unsplash)
Related | OpenAI Is Gearing Up for Another Investment Round That Aims to Turn It Into a $100 Billion Startup
View 0 comments