Tian Yuandong, a Silicon Valley AI guru laid off by Meta: In the AI era, will everyone eventually lose their jobs?

At the end of last month, Meta laid off 600 employees, shocking the world. Meanwhile, Tian Yuandong, a well - known figure in the AI field, officially announced that he was among the laid - off employees.

Tian Yuandong was the former director of AI research at Meta. During his tenure at Meta, he led many important research achievements. According to his Google Scholar profile, the total citation count of his papers has exceeded 18,000.

After being laid off from the top - tier artificial intelligence research department, this AI scientist has received job offers from the world's most powerful technology companies, including OpenAI and Google. He is truly a "genius".

This staff reduction within major AI companies and the departure of prominent figures have also attracted a series of attention from the industry.

Recently, Tian Yuandong was interviewed by the "Silicon Valley 101" program. In addition to discussing the lay - offs, they also talked about the major trends in the AI industry.

He mentioned, "As the degree of automation gets higher and higher, one day everyone will be out of work."

Thanks to AI, tasks that originally required hundreds or even thousands of people can now be accomplished with high - level automation and the application of AI Agents, without the need for as many people.

"In the future, there will be fewer people working on AI, while more people will use AI as a tool to explore other fields."

However, at the same time, Tian Yuandong stated that the core value of human beings still cannot be replaced by AI.

AI lacks high - level human insights, human knowledge, and unique perspectives on problems.

"Throughout the long history of humanity, there have been various scientists with unique ways of thinking. They didn't read as many books or have as much data as others, yet they were able to discover new theorems, new proofs, and new inventions."

Tian Yuandong once gave an excellent explanation at Hundun about the relationship between humans and AI and the speculation about the ultimate outcome of future society.

The general trend of AI will reshape everyone's position. Humans must seek their "uniqueness" - this is Tian Yuandong's most important reminder!

The following are excerpts from the wonderful content.

In the AI era, we need to bid farewell to the "small - town exam - taker" model

Besides conducting research, I'm also a science - fiction writer. I'm thinking about: We're so obsessed with "scaling up large models" now. What's the end of this?

Before the emergence of AI, to accomplish a task, we usually recruited a large number of people to form a team. Each person in the team needed 20 years of education to become competent, and there was a lot of randomness and uncertainty in this process.

However, if one day AI becomes extremely cheap, it will sweep across the world with extremely low costs, high stability, and never - tiring characteristics.

In this situation, what will the market look like? How can individuals survive? How can companies become better?

Currently, the essence of AI is: It has absorbed a large amount of past problem - solving experience. They are very much like "small - town exam - takers" who are very proficient in solving problems.

For problems with known answers or known problem - solving logics, through a large amount of training, AI can find ways and logics to solve problems and do better than humans. It will quickly find possible search paths and try them one by one until it finds the correct logic or answer.

From this perspective, it is believed that AI will completely replace the process of repetitive labor in the future.

Conversely, for us humans, the purpose of solving problems is not just to solve them proficiently. Instead, we hope that through this process, people can learn the more essential logic of problems, better ways of thinking, and cultivate interests, enabling people to reach a higher level than "college - entrance - exam takers" and be able to discover new things.

If a person is only immersed in the "problem - solving" process, it's easy for them to be replaced by AI because the operating mode of AI is to simulate the thinking of an exam - taker.

The AI flood and the human islands: New ecological niches

We can imagine the whole world as many mountains, with a person on each mountain, representing a professional field. When AI arrives, it's like a great flood.

This great flood submerges those "peaks" that are not very high - because the supply of AI is infinite, and it has no scarcity. All jobs and fields that can be replaced by AI below this "sea - level" are completely submerged.

What remains are the fields where people work above the sea - level. These jobs are relatively more creative, or the data is relatively scarce. In this situation, we will see individual islands. On each island, there is an expert in the field who may use their domain knowledge to make good use of AI.

This is likely to happen in the next decade. It will gradually penetrate into every field.

Now, we may still be at home, watching the press conferences of OpenAI or other companies and being amazed. But this is still an early stage. I think after 2025 and 2026, AI technology will gradually penetrate into every field, and people in each field will use AI to revolutionize the entire work process. Only then will this "transformation" truly occur.

Speculation about the ultimate outcome of future society: "Think and get it"

If this process continues, what will ultimately happen? This may be a more science - fiction or fantastical idea: If one day our world is filled with AI, a world where "think and get it" may emerge.

As soon as you think of something, it will immediately appear in front of you. Because the intermediate execution process has been replaced by AI. AI will break down your goals into sub - goals and do its best. In the terms of immortal - cultivation novels, it's called "speak and the law follows" - as soon as you say something, it becomes a reality.

In this situation, the value of repetitive labor will decline significantly. If a task is repetitive labor, then there must be a lot of data sets, and AI can learn from them. So, we will definitely use AI to do it instead of humans.

In this situation, how do we redefine the value of humans? What kind of ecological niche should each individual in society find in this new world? In the next decade, the entire ecosystem will undergo a major reshuffle, and it will be a silent one. It's not a grand narrative like climbing Mount Everest, but it will penetrate into everyone's daily life.

Through AI, I've found that many things I wanted to do but couldn't before can now be done very well. After 2025, this accelerating trend will become more and more obvious. People will find that if you don't use this tool, your efficiency will be far lower than those who use AI.

Uniqueness: The irreplaceable value of humans

If so, what should each of us do?

I think each person will ultimately find their own ecological niche and their own uniqueness. There is no answer to what this "uniqueness" is. Everyone has to find it on their own.

For me, I like to write novels and put some wild ideas in them. This is a unique way of expression for me. Eventually, everyone may find their own interesting and unique ways in the new paradigm.

Just like Liu Bang, he knew he was not as good as Xiao He, Han Xin, and Zhang Liang, but he was able to stand out because he knew his own weaknesses and strengths. In the future, everyone will have many AIs at their disposal, and each AI can do many amazing things. How we can become the "generals" or "commanders" of these AIs and make them work together to achieve what we want may be a very important direction.

The efficiency of large models is still lower than that of humans

From the perspective of technological research, we believe that the learning efficiency of current large models is lower than that of humans.

Although large models perform well in repetitive tasks and are improving rapidly, their learning efficiency is still not as good as that of humans. Humans can acquire in - depth knowledge with very little data. That's why when you ask an AI model some in - depth questions, it often gives some broad, general logics and solutions, but it won't give you very specific, customized in - depth solutions.

There are still many subtle things that machines can't do now.

Therefore, in the future, each person will have their own personalized and creative labor. This kind of labor will generate limited, highly specialized data that is constantly changing. These are things that AI still can't do. They can't accurately understand your thoughts or solve everyone's problems precisely.

This kind of personalized, creative labor should not be replaced by AI. However, repetitive labor will sooner or later be replaced by AI.

That's why I call it "uniqueness".

You have your own unique data, unique way of thinking, and unique experience. These experiences cannot be replaced by large models, and this may be the future value of humans.

Finally, we all need to think about a question: How can we make good use of each person's "data island"? Through this way, can we have our own unique world and unique career in the new ecological environment?

This article is from the WeChat official account "Hundun University" (ID: hundun - university), author: Hundun Academy. It is published by 36Kr with authorization.