OpenAI's Vice President of Engineering, Srinivas Narayanan, shared the importance of AI reasoning (Reasoning) in an interview on 2/12. He mentioned that reasoning enables AI to solve more complex problems, similar to how humans need to think deeply to find solutions when faced with challenges. He emphasized that if AI can develop stronger reasoning capabilities, it can handle more complex tasks with high economic value, aligning with OpenAI's pursuit of general artificial intelligence (AGI).
The era of AI agents is coming, Operator and Deep Research join forces to attack together
Narayanan said that OpenAI recently launched two AI agents, namely Operator and Deep Research, to help users complete tasks more efficiently.
Operator: Able to perform operations in the browser, such as automatic booking, form filling, etc.
Deep Research: Help users conduct more in-depth information research, such as analyzing financial reports, industry trends, etc., and quickly summarize reports.
He believes that AI agents like this will change the operating mode of enterprises, enabling AI to do more than just answer questions, but truly "get things done," significantly improving productivity.
AI inference from medical to financial, all-round development
Narayanan introduces the application of AI reasoning technology in various industries:
Medical field: Medical practitioners use AI to analyze patient data and improve treatment effectiveness. National laboratories use AI to study the correlation between gene mutations and rare diseases.
Biotechnology Industry: AI helps pharmaceutical companies predict clinical trial results and accelerate new drug development.
Financial and Tax Analysis: AI is applied to financial data organization and tax calculation to improve decision-making efficiency.
These applications show that the reasoning ability of AI has begun to bring tangible value to enterprises.
DeepSeek is rising, and the cost of inference models continues to decline
Recently, Chinese AI company DeepSeek launched the R1 inference model, which is said to have a training cost of only a few million U.S. dollars, demonstrating that the cost of AI models is rapidly decreasing. In response to this, Narayanan stated that this is a trend in technological development, and OpenAI is also continuously reducing model costs, for example, the operating cost of GPT-4o has been reduced by 150 times compared to earlier versions.
He emphasized that AI will become cheaper and more popular in the future, and enterprises will also find it easier to use AI to solve various problems.
OpenAI is still discussing whether to open source internally.
When asked if OpenAI will open source its own AI models like O1, O3, etc., Narayanan said the company is still discussing internally. He pointed out that enterprise customers are not only concerned about whether AI is open source, but also about whether AI is easy to use, cost-effective, customizable, and other practical issues.
Currently, OpenAI still prioritizes ensuring model safety and will adjust its open-source strategy according to the situation in the future.
Enterprise AI applications and future development focus on customization and automation
For enterprise needs, Narayanan said OpenAI is developing more customized AI solutions, such as enhancing the Reinforcement Fine-Tuning technology, allowing companies to train dedicated AI models with very little data, requiring only a few dozen examples.
He believes this will change the way enterprises implement AI, allowing different industries such as healthcare, law, and retail to tailor exclusive AI solutions to their own needs.
AI will impact the world like electricity, requiring strong infrastructure
Narayanan mentioned that OpenAI is collaborating with major enterprises to establish AI computing infrastructure, such as the "Stargate" project. He believes that AI will bring a similar impact to the invention of electricity, and future AI models will not only be smarter, but also cheaper and more easily applicable to various industries.
OpenAI has wide internal applications
Narayanan also revealed to the audience present that OpenAI itself also extensively uses AI technology to improve internal efficiency:
Engineering Department: Use AI as a code assistant to help with error correction and code optimization.
Finance and Legal: Using AI to analyze contract terms and reduce potential risks.
Customer Support: AI automatically replies to customer inquiries to improve service efficiency.
(Responding to the impact of DeepSeek on gate! OpenAI updates Model Spec: Emphasizing customization and freedom of thought, releasing a public licensed version of )
This article OpenAI Vice President of Engineering Narayanan: Customization is the key to AI introduction into enterprises, and o1, o3 models are expected to be open source first appeared on ChainNews ABMedia.
The content is for reference only, not a solicitation or offer. No investment, tax, or legal advice provided. See Disclaimer for more risks disclosure.
OpenAI Vice President of Engineering Narayanan: Customization is key to AI adoption by enterprises, o1, o3 models are expected to be Open Source
OpenAI's Vice President of Engineering, Srinivas Narayanan, shared the importance of AI reasoning (Reasoning) in an interview on 2/12. He mentioned that reasoning enables AI to solve more complex problems, similar to how humans need to think deeply to find solutions when faced with challenges. He emphasized that if AI can develop stronger reasoning capabilities, it can handle more complex tasks with high economic value, aligning with OpenAI's pursuit of general artificial intelligence (AGI).
The era of AI agents is coming, Operator and Deep Research join forces to attack together
Narayanan said that OpenAI recently launched two AI agents, namely Operator and Deep Research, to help users complete tasks more efficiently.
Operator: Able to perform operations in the browser, such as automatic booking, form filling, etc.
Deep Research: Help users conduct more in-depth information research, such as analyzing financial reports, industry trends, etc., and quickly summarize reports.
He believes that AI agents like this will change the operating mode of enterprises, enabling AI to do more than just answer questions, but truly "get things done," significantly improving productivity.
AI inference from medical to financial, all-round development
Narayanan introduces the application of AI reasoning technology in various industries:
Medical field: Medical practitioners use AI to analyze patient data and improve treatment effectiveness. National laboratories use AI to study the correlation between gene mutations and rare diseases.
Biotechnology Industry: AI helps pharmaceutical companies predict clinical trial results and accelerate new drug development.
Financial and Tax Analysis: AI is applied to financial data organization and tax calculation to improve decision-making efficiency.
These applications show that the reasoning ability of AI has begun to bring tangible value to enterprises.
DeepSeek is rising, and the cost of inference models continues to decline
Recently, Chinese AI company DeepSeek launched the R1 inference model, which is said to have a training cost of only a few million U.S. dollars, demonstrating that the cost of AI models is rapidly decreasing. In response to this, Narayanan stated that this is a trend in technological development, and OpenAI is also continuously reducing model costs, for example, the operating cost of GPT-4o has been reduced by 150 times compared to earlier versions.
He emphasized that AI will become cheaper and more popular in the future, and enterprises will also find it easier to use AI to solve various problems.
OpenAI is still discussing whether to open source internally.
When asked if OpenAI will open source its own AI models like O1, O3, etc., Narayanan said the company is still discussing internally. He pointed out that enterprise customers are not only concerned about whether AI is open source, but also about whether AI is easy to use, cost-effective, customizable, and other practical issues.
Currently, OpenAI still prioritizes ensuring model safety and will adjust its open-source strategy according to the situation in the future.
Enterprise AI applications and future development focus on customization and automation
For enterprise needs, Narayanan said OpenAI is developing more customized AI solutions, such as enhancing the Reinforcement Fine-Tuning technology, allowing companies to train dedicated AI models with very little data, requiring only a few dozen examples.
He believes this will change the way enterprises implement AI, allowing different industries such as healthcare, law, and retail to tailor exclusive AI solutions to their own needs.
AI will impact the world like electricity, requiring strong infrastructure
Narayanan mentioned that OpenAI is collaborating with major enterprises to establish AI computing infrastructure, such as the "Stargate" project. He believes that AI will bring a similar impact to the invention of electricity, and future AI models will not only be smarter, but also cheaper and more easily applicable to various industries.
OpenAI has wide internal applications
Narayanan also revealed to the audience present that OpenAI itself also extensively uses AI technology to improve internal efficiency:
Engineering Department: Use AI as a code assistant to help with error correction and code optimization.
Finance and Legal: Using AI to analyze contract terms and reduce potential risks.
Customer Support: AI automatically replies to customer inquiries to improve service efficiency.
(Responding to the impact of DeepSeek on gate! OpenAI updates Model Spec: Emphasizing customization and freedom of thought, releasing a public licensed version of )
This article OpenAI Vice President of Engineering Narayanan: Customization is the key to AI introduction into enterprises, and o1, o3 models are expected to be open source first appeared on ChainNews ABMedia.