OpenAI is quietly developing a mysterious project codenamed "Strawberry", whose capabilities may have reached doctoral level

Written by:Samuel Nwite

Compiled by: MetaverseHub

To stay ahead in the rapidly evolving field of artificial intelligence,OpenAI 正在秘密研发一种新的 AI 模型，代号为「草莓」.

The Microsoft-backed startup, best known for its ChatGPT product, is working to demonstrate the advanced reasoning capabilities of its model, which could be a major leap forward for AI, according to internal documents reviewed by Reuters and a person familiar with the matter.

01. A look inside the Strawberry project

The OpenAI team is deep into the Strawberry project, according to a recent internal document seen by Reuters in May. While the document's exact timeline is unclear, it outlines OpenAI's plans to use Strawberry for advanced artificial intelligence research.

The project, described as "in progress," has been shrouded in secrecy even within the company.The goal of Strawberry is to enable AI to not only generate answers, and can autonomously and reliably browse the internet, conducting what OpenAI calls “deep research.”

“This is something that has not been possible with AI models to date,” the source said, noting the project’s ambition.

Asked about Strawberry and the details in the report, an OpenAI spokesperson said in a statement: “We want our AI models to see and understand the world as we do.Continuous research on new AI capabilities is a common practice in the industry, our shared belief is that over time the reasoning capabilities of these systems will continue to improve.”

But the spokesperson did not directly answer the question about "strawberries".

02. From Q to the new era of "Strawberry" reasoning

It is reported that "Strawberry" is the successor to a previous project called Q.

According to two people familiar with the matter, Q is considered a breakthrough within OpenAI because it can answer complex scientific and mathematical questions that go beyond the capabilities of most current commercial models.

At an internal all-hands meeting this year, OpenAI presented a research project that demonstrated new human-like reasoning capabilities, Bloomberg reported.

While Reuters could not confirm whether the project on display was Strawberry, it coincides with the company's ongoing efforts to enhance the reasoning capabilities of artificial intelligence.

OpenAI CEO Altman has emphasized the importance of reasoning in AI, saying earlier this year, “The most important areas of progress will revolve around reasoning abilities.”.

03. Challenges of AI Reasoning

The researchers believe thatImproving the reasoning capabilities of AI models is key to achieving human or superhuman-level intelligenceWhile large language models are effective at summarizing text and writing essays, they often fail on commonsense questions and logical tasks, leading to so-called “hallucinations,” or generating false information.

According to AI researchers,Reasoning involves the AI’s ability to plan, understand the physical world, and solve multi-step problems.

OpenAI’s Strawberry project aims to overcome these challenges by employing a specialized post-training process, which involves fine-tuning AI models after they have been pre-trained on massive datasets.

According to a person familiar with the matter, Strawberry's approach has similarities to Stanford University's "Self-Taught Reasoning" (STaR), which allows artificial intelligence models to iteratively create their own training data, potentially enabling them to reach higher levels of intelligence.

One of the creators of STaR, StanfordXiaobai NavigationUniversity professor Noah Goodman commented: "I think it's both exciting and terrifying... If things continue to go in this direction, we as humans have some serious things to think about."

04. Long-term mission planning and autonomous research

One of the ambitious goals of the Strawberry project is to be able to perform long-duration missions (LHT), which requires AI to plan and execute a series of actions over a longer period of time.

Internal documents show that OpenAI is training and evaluating models on a "deep research" dataset to achieve these capabilities.

While the specifics of the dataset and how long it will be extended remain undisclosed, the goals are clear:Enable AI to conduct research autonomously with the help of Computer Utilization Agents (CUA), and take action based on the findings.

05. The fiercely competitive AI industry

In terms of enhancing artificial intelligence reasoning capabilitiesOpenAI is not alone. Large technology companies such as Google, Meta and Microsoft, as well as numerous academic labs, are also exploring various techniques to improve the reasoning capabilities of artificial intelligence.

However,Opinions differ on whether large language models can incorporate long-term planning and advanced reasoning in predictionsYann LeCun, a pioneer of modern AI at Meta, has often expressed doubts about whether large language models (LLMs) can achieve human-like reasoning.

"Strawberry" represents an important part of OpenAI's strategy.Designed to address the limitations of current AI modelsBy developing more advanced reasoning capabilities, OpenAI aims to unlock new possibilities for artificial intelligence, from scientific discovery to the creation of new software applications.

At the same time, the company has been signaling to developers and partners that it is about to release technology with significantly enhanced reasoning capabilities.

The development of Strawberry includes post-training methods such as fine-tuning, which involves human feedback and an iterative learning process. These techniques are designed to refine AI models and improve their performance on specific tasks.

Advances made through Strawberry technology could redefine what AI is capable of, and set new standards for what these models can achieve.

While the road ahead is challenging, the potential rewards are enormous, heralding a new era of smart, autonomous AI systems.

In the words of an OpenAI spokesperson, “We want our AI models to see and understand the world as we do. If Strawberry is successful, we will be one step closer to achieving that vision.”

OpenAI introduces a five-level system to track its progress towards achieving artificial general intelligence (AGI)These levels cover different levels of AI capabilities, from Level 1, which represents current conversational AI, to Level 5, which envisions being able to manage and execute work across an entire organization.

Here are the five levels of artificial intelligence defined by OpenAI:

1. Chatbots: AI with conversational language

2. Reasoners: Possessing human-level problem-solving abilities

3. Agents: Systems that can take action

4. Innovator: AI that can help invent and create

5. Organizer: Able to complete organizational work

OpenAI believes that Strawberry is approaching Level 2, which involves problem solving similar to a PhD level without the help of tools.The framework aims to provide a structured approach to understanding and developing artificial intelligence systems that can ultimately surpass human intelligence..

The article comes from the Internet:OpenAI is quietly developing a mysterious project codenamed "Strawberry", whose capabilities may have reached doctoral level

related suggestion: cryptocurrency: A revolutionary feast for optimists, grassroots can also change their fate

It is this chaotic, noisy and often misguided technological revolution that stems from people's strong desire to seek independent control over affairs. Written by: Matti Translated by: BitpushNews Yanan People have different views on cryptocurrency: some regard it as a luxury for nerds, some see it as a symbol of monetary anarchism, and some see it as a delusion…