Open AI's new model, o1, can think like a human?

At the time of the valuation of up to $150 billion, OpenAI released a big move, and the "strawberry" model that has been preheated for a long time finally came. OpenAI said that the new model represents a new level of AI ability in reasoning ability, designed to solve complex reasoning problems, training the model to spend more time thinking before responding, similar to the way humans think, the model can solve science, programming and mathematics and other more complex tasks, and than the previous GPT-4o significantly improved.

Traditional model training emphasizes on investing a lot of computational resources in training. The new Open AI model o1 represents a new AI model development paradigm. But to say that it can "think like a human" is a relatively complex statement that requires careful evaluation. The advent of the new model 01 has also triggered widespread discussion. The first is the reasoning ability and application potential of the 01 model, and the o1 model shows excellent ability on complex reasoning tasks. In the 2024 AIME Math exam, o1 successfully solved 74 percent of the questions, far surpassing GPT-4o's 12 percent. In GPQA tests covering difficult chemistry, physics and biology, o1 outperformed even many human experts. o1 uses self-play reinforcement learning, which simulates the slow thinking style of humans and is able to solve problems in complex environments under pressure. This allows o1 to disassemble the problem more deeply, perform chain reasoning, and finally arrive at a solution. Targeted at professional groups such as scientific researchers, programmers, and engineers, o1's reasoning abilities are particularly prominent. These users can use o1 to perform complex data analysis, generate mathematical formulas, or solve multi-step programming tasks. For example, biologists can use o1 to annotate cell sequencing data, physicists can generate complex mathematical formulas needed for quantum optics, and software developers can build and execute complex workflows.

The second is the limitation, the o1 model is relatively expensive to use. o1-preview is priced at $15 per million of input tokens and $60 per million of output tokens, a significant increase compared to GPT-4o. At present, o1 model is still a pure text version model, unable to browse real-time web pages, upload files and pictures, lack a wide range of world knowledge. In addition, in some use cases, o1 may have slower reasoning abilities and take longer to answer questions. At the same time, "thinking like a human being" is a complex and ambiguous concept that involves multiple levels of cognition, emotion, and consciousness. At present, AI models have made significant progress in complex reasoning and problem solving, but there are still huge gaps between them and humans in emotional understanding, creativity, intuition, and so on. Although o1 models excel at complex reasoning, they still rely on large amounts of data and algorithms for training and optimization, and lack autonomous awareness and subjective experience. Therefore, it may be too much to say that o1 can "think like a human."

In addition, the impact of Open AI's new model o1 is also multifaceted, although the application of o1 model will help improve human work efficiency. In tasks that require complex reasoning and problem solving skills, o1 can be a human's right-hand man, helping humans complete tasks faster and improve the quality of their work. However, with the continuous development and application of AI technology, the o1 model may also raise some ethical and privacy issues. For example, how do you ensure that o1's reasoning processes are ethical and legal? How to protect users' privacy and data security? These issues need to be addressed by AI companies, policymakers, and all sectors of society.

To sum up, the advent of the new Open AI model o1 is an important progress in the field of artificial intelligence, and its excellent reasoning ability and wide application potential inject new vitality into the development of AI technology. However, the high cost and market positioning also limit its popularity, which will require OpenAI and related organizations to continue to explore and optimize in the future.