GuShiio.com OpenAI is making every effort to deploy AI Agents, and these thousands of agents will eventually be widely used and become virtual colleagues in various business scenarios.
Author: Sam Altman
Compiled by: AIGC Open Community
At 5 a.m. today, OpenAI co-founder and CEO Sam Altman posted an in-depth article “Three Observations” on his personal blog.
Three observations are mainly made about the AI world. The intelligence level of the AI model is roughly equal to the logarithm of the resources used to train and run it;
The cost of using a specific level of AI drops approximately 10 times every 12 months, while lower prices lead to more use; the social and economic value created by linearly growing intelligence grows super-exponentially.
The article also specifically emphasized that OpenAI is making every effort to deploy AI Agents, and these thousands of agents will eventually be widely used and become virtual colleagues in various business scenarios. The performance of junior agents may be worse, but they will still penetrate into various fields to reshape the world economy.
The following is the original content:
Our mission is to ensure that General Purpose AI (AGI) benefits all mankind.
The system that started pointing to AGI was already beginning to emerge, so we felt the need to understand the moment we were in. AGI is a vaguely defined term, but generally we define it as a system capable of solving increasingly complex problems at the human level in many fields.
Human beings are tool makers and are born with the inherent motivation to explore and create, which makes the world a better and better place. Each generation continues to build on the discoveries of its predecessors, creating more powerful tools such as power, transistors, computers, the Internet, and the upcoming AGI.
In a sense, AGI is just another tool on the ladder of progress that mankind has jointly built. But in another sense, the emergence of AGI marks a new beginning, and it’s hard not to feel that this time is really different. The economic growth before us looks amazing, and we can even imagine a world that can cure all diseases, allow people more time to spend with their families, and unleash their creativity to the fullest.
Perhaps within ten years, everyone on earth will be more capable than the most influential people today.
We continue to witness rapid progress in AI development. Here are three observations about AI economics:
1. The intelligence level of an AI model is roughly equal to the logarithm of the resources used to train and run it. These resources mainly include training computing capabilities, data and reasoning computing capabilities. It seems that as long as you are willing to invest any amount of money, you can get sustained and predictable returns. The law of scale for predicting such gains is accurate on multiple orders of magnitude.
2. The cost of using a specific level of AI drops approximately 10 times every 12 months, while lower prices lead to more usage. You can see this in the change in token costs from GPT-4 in early 2023 to GPT-4o in mid-2024, where the price per token has dropped by about 150 times over a year and a half.
Moore’s Law has changed the world by doubling performance every 18 months; the cost of AI is falling even faster than that.
3. The social and economic value created by linearly growing intelligence grows exponentially. This result shows that we see no reason why exponentially growing investment will stop in the near future.
If these three observations continue to be true, their impact on society will be huge.
We are now starting to roll out AI Agents that will eventually be as widely used as virtual colleagues.
Let’s take software engineering agents as an example, which we think are particularly important. Imagine that this kind of Agent will eventually be able to complete tasks that most top-level corporate software engineers with several years of experience can complete (tasks up to a few days). It may not generate the biggest new ideas, require a lot of human supervision and guidance, and perform well in some aspects and may perform poorly in others.
But you can still imagine him as a real but relatively rudimentary virtual colleague. Now imagine having 1000 such Agents. Or a million. Imagine having such an Agent in every field of knowledge work.
In a sense, AI may be economically similar to transistors, a major scientific discovery that can be scaled up on a large scale, permeating every corner of the economy. We won’t focus too much on transistors or transistor companies, but we will expect our computers, TVs, cars, toys, etc. to play a miraculous role.
The world doesn’t change overnight, it never has. In the short term, life will remain roughly the same, and people in 2025 will still spend most of their time as they did in 2024. We still fall in love, start a family, argue online, go hiking, etc.
But the future will hit us in a way that cannot be ignored, and the long-term changes in society and economy will be huge. We will find new things to do, new ways to help each other, and new ways to compete, but these may be very different from today’s jobs.
Autonomy, will and determination can become extremely important. Making the right decisions about what to do and how to navigate a changing world will be of great value; resilience and adaptability will be skills worth developing. AGI will be the strongest lever of human will, and it will enable individuals to have a greater impact than ever before, not less.
We expect the impact of AGI to be uneven. Although some industries may not change much, scientific progress may be faster than today, and this impact of AGI may outweigh everything else.
The prices of many goods could end up falling sharply (intellectual and energy costs currently limit many things), while the prices of luxury goods and some inherently limited resources such as land could rise more dramatically.
From a technical perspective, the path ahead seems quite clear. But public policy and collective opinion are crucial in how to integrate AGI into society. One reason we launch products early is to give society and technology time to evolve together.
AI will penetrate into all areas of the economy and society, and we expect everything to become smart. Many expect the need to give people more control over technology than in history, including more open source, and accept the trade-off between security and personal empowerment.
While we never want to act recklessly, and there may be some major AGI security-related decisions and restrictions that will be unpopular, in terms of direction, as we get closer to achieving AGI, we believe it is important to tend to empower individuals more. Another possible path we can see is AI being used by authoritarian governments to control their populations through large-scale surveillance and deprivation of autonomy.
Ensuring that the benefits of AGI are widely distributed is crucial. The historical impact of technological progress suggests that most of the indicators we care about (health outcomes, economic prosperity, etc.) will get better on average and in the long run, but increasing equality does not seem to be determined by technology, and achieving this correctly may require new ideas.
In particular, the balance of power between capital and labor may be disrupted, which may require early intervention. We are open to ideas that may sound outlandish, such as allocating a computing budget to everyone on the planet so they can use large amounts of AI, but we can also see many ways to achieve the desired results by continuously reducing the cost of intelligence.
By 2035, anyone should be able to call on intellectual abilities equal to the combined intelligence of all people in 2025; everyone should be able to gain unlimited wisdom and use it in any way they can imagine. Nowadays, there are many talents in the world who don’t have enough resources to fully express themselves. If we can change this, the world will create tremendous value that will bring huge benefits to all of us.
Welcome to join the official social community of Shenchao TechFlow
Telegram subscription group: www.gushiio.com/TechFlowDaily
Official Twitter account: www.gushiio.com/TechFlowPost
Twitter英文账号：https://www.gushiio.com/DeFlow_Intern
