news

OpenAI is at it again.

On the 13th,ChatGPTPosted on social media X, chatbot familyGPT-4oThe model has been updated again, and it is emphasized that this time it will be very different. This model is notOpenAIThe latest API-side GPT-4o 0806 model can also be used by free users, but there is a limit on the number of uses.

Before the official announcement, there were rumors in foreign media that OpenAI would soon announce the "Strawberry" project in the update of ChatGPT.

After the official announcement, some netizens also speculated that this new GPT-4o model might be Strawberry.

These speculations are not groundless.

ChatGPT had previously posted a strawberry emoticon on X, which seemed to be hinting at something.

Open AI's CEO Altman also posted a photo of strawberries in his summer garden on social media, which made people imagine.

So what exactly is this "Strawberry" project?

Last month, OpenAI’s internal team was revealed to be developing a “Strawberry” project.The goal is to enhance the reasoning capabilities of OpenAI’s models and their ability to handle complex scientific and mathematical problems, so that large models can not only generate answers to queries, but also plan ahead in order to autonomously and reliably browse the Internet and conduct what OpenAI defines as “deep research.”。

The predecessor of this "Strawberry" project is the Q* algorithm model, which can solve difficult scientific and mathematical problems. Mathematics is the foundation of the development of generative AI. If the AI ​​model masters mathematical ability, it will have stronger reasoning ability and even be comparable to human intelligence. This is also something that the current large language model cannot do.

Altman once emphasized that the key to the future development of AI will revolve around reasoning ability.

According to people familiar with the matter, the "Strawberry" project includes a special"Post-training"Method, that isAfter the generative AI model has been pre-trained on a large dataset, it is further tuned to improve its performance on a specific task. This is similar to the "Self-Taught Reasoner" (STaR) method developed by Stanford University in 2022.

Noah Goodman, one of the creators of STaR and a professor at Stanford University, once said that STaR can allow AI models to "guide" themselves to a higher level of intelligence by repeatedly creating their own training data. In theory, it can be used to enable language models to achieve intelligence that surpasses that of humans.

This is in line with the "reasoning ability" that OpenAI wants. In addition, OpenAI also announced a five-level roadmap for the future development of AI on the 11th:

According to OpenAI's conjecture, AI will go through five stages in the future:

Level 1: Chatbots, artificial intelligence with conversational language

Level 2: Reasoner, human-level problem solving

Level 3: Agents, systems that can take action

Level 4: Innovators, AI that can assist with invention

Level 5: Organizer, AI that can complete organizational work

According to the information collected so far, the "Strawberry" project is very likely to be the key to helping OpenAI achieve second-level AI.

If OpenAI succeeds, Strawberry will likely redefine the capabilities of AI, enabling it to make major scientific discoveries, develop new software applications, and perform complex tasks autonomously, putting humans within reach of the next generation of AI.AGIAnother step forward.