Sam Altman: Three Observations on AI Economics

This article is from: The Paper News

Ultraman: Three Observations on AI Economics

On February 10th, OpenAI CEO Sam Altman shared his three observations on the economics of artificial intelligence in his personal blog post, 'Three Observations.' He believes that the scaling law of artificial intelligence will continue to be effective, and there is no reason to stop investing in AI's exponential growth in the near future. Achieving increased equality is not determined by technology alone, and it may require new ideas, especially as the balance of power between capital and labor is easily disrupted, which may necessitate early intervention.

“Our mission is to ensure that AGI (Artificial General Intelligence) benefits all of humanity,” Ultraman said. Humans are creators of tools, with an inherent drive to understand and create. Electricity, transistors, computers, the internet, and the upcoming AGI, each new generation of technology builds upon the discoveries of the previous generations, creating more powerful tools. In a sense, AGI is just another tool in the scaffolding of human progress that we collectively build. In another sense, this is a beginning, and it's hard not to say 'this time is different.'

Artificial intelligence continues to develop rapidly. Ultraman shared three observations on the economics of artificial intelligence.

First, the intelligence of AI large models is roughly equivalent to the resources used for training and running them, which are mainly training computation, data, and inference computation. It seems that you can spend any amount of money and get continuous and predictable returns. The law of scale for predictions is accurate across many orders of magnitude.

Second, the cost of using AI at a specific level drops by about 10 times every 12 months, and lower prices will bring more applications. This can be seen from the token cost of GPT-4o, which dropped by about 150 times during the period from the beginning of 2023 to the middle of 2024, despite Moore's Law doubling the world every 18 months, this is incredibly powerful.

Thirdly, the socio-economic value of intellectual linear growth is essentially super-exponential. One consequence of this is that there is no reason to stop investing in exponential growth in the near future.

Ultraman said that if these three observations continue to hold, the impact on society will be significant. Now, the AI ​​intelligent agents are being launched, and these agents will eventually make people feel like virtual colleagues. For example, software engineering intelligent agents will eventually be able to handle most of the work that software engineers with several years of experience in top companies can do. But the intelligent agents will not have the greatest new ideas, and they will need a lot of human supervision and guidance.

"In some ways, artificial intelligence may be as significant a scientific discovery for the economy as the transistor, as it can expand and permeate almost every corner of the economy." Ultraman said that the world will not change suddenly. In the short term, life will remain basically unchanged, and people will spend 2025 in the same way as in 2024, but the future will come to us in an undeniable way, and the long-term changes to society and the economy will be enormous. We will find new things to do, find new ways that are useful to each other, find new ways of competition, but they may not look much like today's jobs."

Agility, willpower, and determination are more valuable. Making the right decisions on what to do and figuring out how to navigate in a constantly changing world will be of great value. Resilience and adaptability will be beneficial skills to cultivate. AGI will be the greatest lever ever to impact human will, enabling individuals to have a greater influence than ever before.

Technically, the road ahead of us looks quite clear. The impact of AGI is expected to be uneven, with some industries experiencing minimal changes, but scientific progress could be much faster than today. It is crucial to consider how to integrate AGI into public policies and collective opinions. Many people expect to have more technological control than ever before, including more open source and a balance between security and personal authorization. 'While we don't want to act recklessly, there may be significant decisions and restrictions related to AGI safety that will be unwelcome. However, in terms of direction, as we get closer to achieving AGI, we believe that empowering individuals is important.'

The historical impact of technological progress suggests that most indicators we care about, such as health outcomes and economic prosperity, tend to improve on average and in the long run. However, increasing equality does not seem to be determined by technology, and achieving this may require new ideas. In particular, the balance of power between capital and labor is easily disrupted, which may require early intervention.

"We are open to some strange ideas, such as providing some 'computational budget' so that everyone on Earth can fully use artificial intelligence. We can also see many ways to achieve the expected effect as long as the cost of intelligence continues to decrease." Ultraman said that by 2035, anyone could mobilize intelligence equivalent to that of all humanity in 2025, and everyone should have unlimited talent to guide them to what they can imagine. A lot of talents now don't have the resources to fully express themselves. If we change this situation, the world's creative output will bring huge benefits to everyone.