What is Open AI's mysterious project "Strawberry"?
Is OpenAI tinkering with mysterious “black technology” again? On the 12th, according to media quoting people familiar with the matter, OpenAI, the leader in generative AI, is developing a new AI big model project - “Strawberry” (Strawberry), and this “strawberry” is very mysterious and probably related to improving the reasoning ability of AI models.
The Development of "Strawberry"
An internal document from OpenAI in May shows that the “Strawberry” project being developed by OpenAI's internal team aims to enhance the reasoning ability 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. This will enable them to browse the Internet autonomously and reliably and carry out “in-depth research” as defined by OpenAI.
Q* was first revealed in an internal letter from OpenAI at the end of last year, and CEO Altman was fired at the time because of this Q* project. Some OpenAI insiders pointed out that Q* may be a breakthrough in OpenAI's quest for general artificial intelligence (AGI). The speed at which it is developing is shocking and worried that the rapid development of AI may threaten human security.
The Roadmap for Future AI Development
OpenAI also announced a five-level roadmap for future AI development. According to OpenAI's conjecture, AI will go through 5 stages in the future:
- First level: chatbots, artificial intelligence with conversational language
- Level 2: Reasoner, human-level problem solver
- Third level: agents, systems that can act
- Level 4: Innovators, artificial intelligence that can help invent
- Fifth level: organizer, artificial intelligence that can complete the work of the organization
The Future of "Strawberry"
According to media reports, an OpenAI executive said that the AI model is currently at the first level, but it is expected that it will soon reach the second level, the inference. OpenAI is currently working to achieve doctoral-level intelligence on specific tasks, which is expected to happen within a year or a year and a half. Another feature that OpenAI focuses on is improving the ability of large language models to perform long-time tasks (Long-Horizon Tasks, LHT for short).
If OpenAI succeeds, the “Strawberry” project is likely to redefine AI's ability to make major scientific discoveries, develop new software applications, and perform complex tasks independently, then humans will be one step further away from AGI.
