AGI: How Artificial General Intelligence is changing the way we think about the future

Literal translation material from Sam Altman's blog: 

Sam Altman:

Our mission is to make AGI (Artificial General Intelligence) benefit all of humanity.

Systems are already emerging that indicate AGI is approaching, and we think it is important to recognize where we are now. The term AGI is still loosely defined, but generally refers to a system capable of solving increasingly complex human-level problems in a wide variety of domains.

Humans are tool makers, driven by an innate desire to understand and create, and this leads to the world as a whole becoming a better place for us all. Each new generation builds on the discoveries of the previous generation to create even more powerful tools - electricity, the transistor, the computer, the Internet, and now AGI.

Over time, albeit intermittently, the steady march of human innovation has brought levels of prosperity and improved quality of life that once seemed unimaginable.

On the one hand, AGI is just another tool in the endlessly growing "scaffolding" of human progress that we are all building together. On the other hand, it is the beginning of something that is hard not to say "this time it's different": the coming economic growth looks amazing, and we can already envision a world where we will cure all diseases, have much more time for family, and realize our full creative potential.

Perhaps in as little as a decade from now, every person on Earth will be able to achieve more than the most powerful person can today.

We are seeing the development of AI going very fast. Below are three comments on the economic aspects of AI:

"Intelligence" of an AI model is approximately equal to the logarithm of the resources used to train and operate it. These resources mainly include computational power for the training phase, data, and computational power for inference (inferencing). It seems that one can spend almost any amount of money to achieve continuous and predictable improvements: the scaling laws that describe this turn out to be accurate within many orders of magnitude.

The cost of using a certain level of AI falls by a factor of about 10 every 12 months, and a drop in price leads to a significant increase in usage. This can be seen in the value of tokens in GPT-4 in early 2023 and GPT-4o by mid-2024, when the price per token dropped about 150 times over that period. Moore's Law changed the world by doubling capacity every 18 months; what we're seeing now is incredibly more powerful.

The socio-economic value of linearly growing "intelligence" has a super-exponential character. It follows that we see no reason why the exponential growth of investments should stop in the near future.

If these three factors persist, the impact on society will be significant.

We're already starting to bring AI "agents" to market that will eventually feel like virtual coworkers.

Let's imagine a "programmer agent", which we expect to be particularly important. Let's imagine that, over time, he will be able to do most of the tasks that a programmer with several years of experience at a leading company does when it comes to projects of a couple days. He won't have the most ambitious new ideas, he'll need a lot of human supervision and guidance, and he'll be great at some things and surprisingly weak at others.

But still imagine him as a real, albeit relatively "junior", virtual colleague. Now imagine there are a thousand such agents. Or a million. And then imagine that there are such agents in every field of mental labor.

In some ways, AI may be like the transistor in economic terms - a major scientific discovery that scales well and gradually penetrates almost every corner of the economy. We don't often think about transistors and the companies that make them, and the benefits are very widely distributed. Nevertheless, we expect our computers, televisions, cars, toys, and more to work wonders.

The world doesn't change overnight - it's never happened before. In the short term, life will mostly go on as it has in the past, and in 2025, people will spend their time roughly as they did in 2024. We will continue to fall in love, start families, argue online, go camping, and so on.

But the future will be approaching so obviously that it will be impossible to ignore, and the long-term changes in society and the economy will be enormous. We will find new occupations, new ways of being useful to each other and new ways of competing, but they may not look much like current occupations.

Initiative, will and tenacity are likely to be highly valued. The ability to determine what to do and to navigate an ever-changing world will be essential; perseverance and adaptability will be important skills. AGI will be the most powerful "lever" for human determination and will enable individuals to influence the world more strongly than ever before, not weaker.

We expect the impact of AGI to be uneven. Some industries will change little, but scientific progress is likely to accelerate much more than it is now; and this impact of AGI may outstrip all other factors.

Over time, the cost of many goods will fall rapidly (at the moment, the price of many things is mainly influenced by the cost of "intelligence" and energy), while the cost of luxury goods and some limited resources, such as land, may rise even more.

Technically, we think the path forward is pretty clear. But public policy and collective opinion on how we should integrate AGI into society is very important; one of the reasons we release products early and often is to give society and technology time to evolve together.

AI will penetrate all areas of the economy and society; we will expect everything around us to become 'smart'. Many of us believe that in the future people will have to give more control over technology than they have (historically) been given, including opening up more source code, and recognizing that the balance between security and personal freedom involves trade-offs.

We never want to act recklessly, and there are likely to be some serious decisions and constraints on AGI security that will not be to everyone's liking. But in general, as we get closer to full AGI, we think it's important to shift towards empowering everyone. The other way we see AI being used by authoritarian governments for total surveillance and suppression of freedom.

It is critical that the benefits of AGI are widely distributed. Historical experience suggests that most indicators (health, economic well-being, etc.) improve on average over time, but inequalities do not resolve themselves, and new ideas are likely to be needed to do so.

In particular, the balance of power between capital and labor could easily be upset, and early intervention may be necessary. We are open to out-of-the-box ideas, like giving everyone on the planet a "computing budget" so that they can make large-scale use of AI. But we also see many ways in which steadily reducing the cost of "intelligence" will itself produce the desired result.

Anyone in 2035 should be able to amass intellectual resources equivalent to those possessed by all of humanity in 2025; everyone should have access to "unlimited genius" that can be applied as soon as it seems possible. There are already a huge number of talented people who don't have enough resources to reach their full potential. If we change this, the creative input that the world will receive will be of enormous benefit to us all.

By using the term AGI here, we want to maximize clarity and do not intend to change or in any way interpret the definitions and processes underlying our relationship with Microsoft. We are confident that we will partner with Microsoft for the long term. This footnote may look strange, but we know that some journalists want to "collect clicks" with provocative headlines, so we thus preemptively stop such nonsense...

---

We stand on the threshold of a new era where everyone can use AGI's achievements to change their lives and society. It is important to remember that the future depends on our joint contribution and responsible approach - and this is just the beginning of the great journey we all have to take.