What’s OpenAI planning for next year after GPT-4? Where is OpenAI’s moat? What is the value of an AI Agent? With many old employees leaving, will OpenAI choose younger people with more conviction and vitality?
On November 4th, OpenAI CEO Sam Altman answered these questions in a podcast called “The Twenty Minute VC”. He explicitly stated that improving reasoning ability has always been OpenAI’s core strategy.
When the podcast host and 21VC founder Harry Stebbings asked Altman what opportunities OpenAI could offer AI entrepreneurs, Altman believed that if AI entrepreneurs persist in solving the problem of insufficient models, their business models will become less competitive with the upgrades of OpenAI models. Entrepreneurs should instead build businesses that can benefit from the strengthening of models, which will be a huge opportunity.
In Altman’s view, the current way of discussing AI is somewhat outdated. Compared to models, systems are a more worthy direction of attention, and next year will be a crucial year for OpenAI’s transition to AI systems.
The following are selected highlights from Stebbings’ conversation with Altman:
OpenAI plans to create a no-code tool
Stebbings: I’ll start the interview with a question from the audience. Will OpenAI focus on releasing more models like GPT-3.5 or training larger and more powerful models in the future?
Altman: We will optimize the models comprehensively, and improving reasoning ability is the core of our current strategy. I believe that powerful reasoning ability will unlock a series of functions we expect, including AI making substantial contributions to scientific research and writing highly complex code, which will greatly promote social development and progress. You can expect continuous and rapid iteration and optimization of the GPT series models, which will be the focus and priority of our future work.
Sam Altman interviewed by 21VC founder Harry Stebbings
Stebbings: Will OpenAI develop no-code tools for non-technical people in the future, allowing them to easily build and expand AI applications?
Altman: Without a doubt, we are steadily progressing towards this goal. Our initial plan is to significantly improve the efficiency of programmers’ work, but in the long run, our goal is to build top-notch no-code tools. Although there are already some no-code solutions in the market, they currently cannot fully meet the needs of creating a complete startup company in a no-code manner.
Stebbings: In which areas of the technical ecosystem will OpenAI expand? Considering that OpenAI may dominate the application layer, would it be a waste of resources for startups to invest heavily in optimizing existing systems? How should founders think about this issue?
Altman: Our goal is to continuously improve our models. If your business is only focused on solving minor shortcomings of existing models, your business model may become less competitive once our models become powerful enough to eliminate these shortcomings. However, if you can build a business that can benefit from the continuous improvement of models, it will be a huge opportunity. Imagine if someone told you that GPT-4 would become exceptionally powerful and capable of accomplishing tasks that currently seem impossible, you would be able to plan and develop your business from a longer-term perspective.
Stebbings: We have discussed with venture capitalist Brad Gerstner the potential impact of OpenAI on certain niche markets. From the perspective of a founder, which companies might be affected by OpenAI, and which companies might be spared? As investors, how should we evaluate this issue?
Altman: Artificial intelligence will create trillions of dollars in value and lead to the emergence of new products and services, making previously impossible or impractical things possible. In some areas, we expect models to be powerful enough to achieve goals effortlessly, while in other areas, this new technology will be further enhanced by building excellent products and services. In the early days, about 95% of startups seemed to bet that the models would not improve, which surprised me, but now I am no longer surprised. When GPT-3.5 was just released, we foresaw the potential of GPT-4 and knew it would be very powerful. So if the tools you build are only to compensate for the shortcomings of models, these shortcomings will become increasingly irrelevant as the models continue to improve. In the past, when models performed poorly, people were more inclined to develop products that compensated for the model’s deficiencies, rather than building revolutionary products like “AI teachers” or “AI medical consultants”. I felt that at that time, 95% of people were betting that the models would not improve, while only 5% believed that the models would get better. Now the situation has reversed, and people understand the speed of improvement and our development direction. This issue is no longer so prominent, but we were very concerned about it because we foresaw that companies striving in that direction (compensating for model deficiencies) might face difficulties.
Stebbings: You have said that “artificial intelligence will create trillions of dollars in value,” and Masayoshi Son (founder and CEO of SoftBank Group) predicted that “AI will create $90 trillion in value annually,” enough to offset what he believes is “the necessary $90 trillion capital expenditure.” What are your thoughts on this?
Altman: I cannot provide an accurate number. Obviously, a large amount of capital expenditure will create tremendous value because that’s how every major technological revolution works, and artificial intelligence is undoubtedly one of them. Next year is a crucial year for us as we enter the era of the next generation AI system. Regarding the development of no-code software agents that you mentioned, I’m not sure how long it will take, as it is currently not realizable. But if you imagine that we can achieve this goal and everyone can easily access a complete set of enterprise-level software, imagine how much economic value it will unleash. If you can maintain the same value output while making it more convenient and cost-effective, it will have a huge impact. I believe we will see more similar examples, including in the healthcare and education sectors, which represent trillions of dollars in markets. If AI can drive new solutions in these areas, the specific numbers are not important, what is important is that it will indeed create incredible value.
Outstanding AI Agents have capabilities beyond human abilities
Stebbings: What role do you think open source will play in the future development of artificial intelligence? How will the discussion on whether to open-source certain models take place within OpenAI?
Altman: Open-source models play a crucial role in the AI ecosystem. There are already some excellent open-source models available. I think it is also essential to provide high-quality services and APIs. In my opinion, it makes sense to offer these elements as a product suite so that people can choose the solutions that best suit their needs.
Stebbings: In addition to open source, we can also provide services to customers through agents. How do you define “Agent”? What is it, and what is it not in your view?
Altman: I think an Agent is a program that can perform long-term tasks with minimal human supervision during the execution process.
Stebbings: Do you think there is a misconception about Agents among people?
Altman: Rather than a misconception, I would say that we have not fully understood the role that Agents will play in the future world. The examples often mentioned are AI Agents helping to book restaurants, such as using OpenTable or directly calling the restaurant. While this can save some time, what I find more exciting is that Agents can do things that humans cannot do, such as simultaneously contacting 300 restaurants to find the most suitable dishes or restaurants that can provide special services. This is almost an impossible task for humans, but if the Agents are all AI, they can handle it in parallel, solving the problem easily. Although this example is simple, it demonstrates the capabilities of Agents that surpass human abilities. What’s more interesting is that Agents can not only help you book restaurants but also work as a very smart senior colleague who can collaborate with you to complete a project. Or it can independently complete a task that would take two days or even two weeks, only contacting you when encountering problems, and ultimately presenting excellent results.
Stebbings: Will this Agent model impact the pricing of SaaS (Software as a Service)? Traditionally, SaaS is priced based on user seats, but now Agents are effectively replacing humans. How do you view the change in future pricing models, especially when AI Agents become a core part of enterprise employees?
Altman: I can only make guesses because we really cannot be sure. I can imagine a scenario where future pricing models are determined based on the computing resources you use, such as whether you need 1 GPU, 10 GPUs, or 100 GPUs. But it’s still too early to say for sure.
