OpenAICEO's First Response: Not Excluding Self Developed Chips | Exclusive to Titanium Media

OpenAICEOSamAltman

On October 17th, US time,At the WSJTechLive closed door meeting, Sam Altman, the "father of ChatGPT" and CEO of OpenAI, participated in a round table conversation for about 50 minutes, discussing topics such as the development direction of AGI (General Artificial Intelligence) and the next generation of GPT, and responding for the first time to some industry questions about the next steps of OpenAI development.

The Titanium Media App recorded relevant content on-site and was also released for the first time on a Chinese platform.

During the conversation, Ultraman responded for the first time to the "chip making" message, stating that OpenAI does not rule out the self-developed chip plan to meet the computational power requirements of the model.

Ultraman believes that AI technology will change the meaning of work and has the potential to change the global economy by improving productivity and efficiency, requiring society to work together to cope with its rapid development. Although ChatGPT is not yet a comprehensive replacement for manual labor, OpenAI is striving to improve its efficiency and popularity.

He emphasized that in the coming decades, AI will further penetrate into people's lives, with the key being abundant and affordable intelligence and energy.

In addition, 38 year old Ultraman has responded to important topics such as OpenAI financing and valuation, AI regulation, and future AI development.

On "Core Making": OpenAI Does Not Exclude Self Developed Chips

The development of generative AI technology represented by ChatGPT is constantly accelerating and is increasingly favored by more consumers and enterprises.

Recently, there has been news that OpenAI is exploring the possibility of self-developed AI chips and has evaluated potential acquisition targets to accelerate the development of customized chips.

In response to this, Ultraman stated for the first time in the roundtable discussion that OpenAI does not rule out the option of self-developed chips.

We are still evaluating whether to adopt customized hardware (chips). We are working hard to determine how to scale up to meet the world's needs. Although we may not develop chips, we are maintaining good cooperation with partners who have achieved outstanding results, "Altman stated.

The ultimate goal is to surpass NVIDIA.

According to the OpenAI official website, the company has started recruiting hardware engineers to evaluate and jointly design AI chips.

Ultraman emphasized that due to the surge in market demand, the "magic of capitalism" may bring more competitors and ultimately alleviate the shortage problem for their competitors. It is reported that Nvidia currently controls over 80% of the global AI computing chip market.

Talking about AI Replacement: ChatGPT cannot fully replace human work yet

Some analysts believe that this technology can not only improve production efficiency, but also promote economic growth.

According to Goldman Sachs data, generative AI may increase the average annual growth rate of productivity in the United States by 1.5 percentage points in the 10 years after its widespread application, thereby increasing the long-term economic growth expectation of US Gross Domestic Product (GDP) from 1.8% to 3.3% (theoretical value).

However, Ultraman stated that ChatGPT is not yet a comprehensive replacement for human work.

I think for a long time, human experts in certain fields will still have an advantage over AI systems. So you may feel that GPT-4 is doing very poorly in that field, and GPT-5 and GPT-6 may also be doing very poorly. You can use GPT for other tasks, which is not a problem, but it is definitely not an expert in that field. It's a bit like different people in the world can achieve an average level, "Altman said, At present, GPT-4 is mainly suitable for the knowledge environment of the public, and perhaps future versions will enhance relevant professional data to help humans complete more tasks.

Ultraman believes that every technological revolution will affect the job market. Every few hundred years, people lose their jobs due to new technologies, changing their production methods, but they will find better jobs. Considering that a generation takes approximately 25 years, after two generations or approximately 50 years, society can adapt to this change.

When Zhao Hejuan, the founder of Titanium Media, talked about the potential impact of AGI on the news industry, Ultraman responded that he saw almost all the helpful aspects, and the feedback we heard from reporters was that it was surprisingly helpful for their work.

Ultraman pointed out that, The society we live in needs to face the rapid changes brought about by new technologies. In two to three generations, we can adapt to these changes, and society can cope with changes in various jobs and markets. People are eager to achieve a sense of achievement in this process and become useful people who contribute to society. This desire to contribute to society is innate and will not disappear. In addition, humans also have the ability to creatively express themselves Force. We should all work together to become the 'architects' who shape the future

Ultraman also stated that OpenAI will not be the only player in the AGI market, as many people will place large model technologies such as GPT on the technological development track, and OpenAI will be one of the participants in building complex AI systems.

In the future, there will always be AI in our lives, which is the way to progress, "Ultraman said. OpenAI has explicitly decided to call its chat robot ChatGPT instead of a person's name, so that people will not confuse the tool with a person's name.

In his view, once AGI technology, including ChatGPT, is available, humans are more suitable for creative work tasks.

Financing: OpenAI requires a large amount of financing to find R&D talents

On October 19th, it was reported that OpenAI was in talks to sell its existing employee shares at a valuation of $86 billion, but OpenAI had not yet determined a specific distribution plan. OpenAI's valuation is second only to ByteDance and SpaceX, and is expected to become the world's largest AI unicorn company.

At the same time, according to earlier disclosure, OpenAI's annual revenue reached $1.3 billion (approximately RMB 9.493 billion) in 2023, with an average monthly revenue of over $100 million, an increase of more than 45 times from last year's full year of $28 million, reaching 4542%, and also 30% higher than the expected annual revenue three months ago. The company's main source of revenue is the sales revenue from ChatGPT personal paid version and large model API authorization.

Regarding this, Ultraman first said 'No', and then he stated that currently OpenAI needs to raise a large amount of funds to find R&D talents and quickly promote innovative research.

We need a lot of funds to complete the OpenAI mission. In our limited time, there must be something more interesting to discuss than our future financing plan. Of course, we need more money, and the future path of AI is not clear, but innovation is expensive. In terms of products, we are looking for R&D talents and rapidly advancing innovation, and we are working hard to promote various applications of AI and collaborate with customers, "Altman said.

It is reported that OpenAI, founded in 2015, was originally a research oriented non-profit organization funded by investors such as Musk, with the aim of building secure AI technology. Subsequently, in 2019, OpenAI transformed into a "cap profit" structure, with even WeSoft, which invested over $10 billion, holding a minority stake in the company. Ultraman has repeatedly expressed that OpenAI does not intend to go public through an IPO.

Ultraman pointed out that as large language models such as GPT become smarter and more capable, OpenAI will require less data for training.

When asked about the challenges that users may encounter when determining whether content is generated by artificial intelligence, OpenAICTO Mira Murati stated that the company is developing technical tools to help detect image sources, and "almost 99% of the information is reliable," but is still under testing.

Talking about hardware: I don't want to compete with smartphones

Recently, it has been reported that Ultraman is in private contact with former Apple Chief Designer JonyIve and Japanese Softbank Group CEO Sun Zhengyi, reportedly planning to create a brand new AI hardware product. Sun Zhengyi has repeatedly stated that he is very optimistic about the development of OpenAI company and generative AI technology.

In the round table dialogue, Ultraman stated that the market needs to have enough new technologies and computing platforms, which can help generate new value. Ultraman hinted that the hardware devices he plans to build mainly revolve around "computing platforms", rather than popular products in fields such as smartphones and humanoid robots.

Smartphones are a great invention, and I have no interest in competing with them. They excel in their field of expertise. Therefore, I believe that what AI brings is new, which makes more hardware possible. I think it's worth taking time to discuss or think about a question: Can we create things that we couldn't have done before before we had computers to think about or understand? If so No, I would feel a bit disappointed Ultraman said.

Ultraman also emphasized, "I don't think we will have a humanoid robot that can end the world

He judged that if the AI hardware direction plan is correct, it will continue to carry out research and development, which is not only related to the development of GPT technology, but also new computing platform hardware that will generate new value for customers.

Talking about AI Security: Overregulation can be catastrophic

AI security and regulatory issues remain one of the topics of concern to the outside world. There are reports that the European Union plans to implement stricter regulations on generative AI models and is considering using a three-level approach to regulate AI models. This aims to address the security risks brought by generative AI by establishing mandatory rules while avoiding excessive regulatory burden on startups.

Ultraman remains optimistic and supportive on the topic of AI regulation.

Ultraman believes that it is difficult for laboratories to ensure technological safety, and technologies such as DeepFakes have a profound impact on society and require some regulation of related technologies. Faced with the prospect of super intelligence, international regulation and a cautious collaborative approach are needed.

But he also pointed out that government regulation needs to recognize the importance of autonomy for start-up AI companies and open source AI technology. Excessive regulation can be catastrophic.

Ultraman emphasized that although technology may not be perfect at the beginning, such as airplanes, over time, through meticulous engineering and sharing of best practices, AGI technology will make society safer.

We should face the reality of technological development together, ensure economic and social interests, and be prepared for the future, "Ultraman said.

(This article premieres the Titanium Media App, author | Lin Zhijia)