Interpreting Ilya's Latest Speech: Learning is the Last Line of Control in the AI Era

“AI"Ultimately, we will possess unpredictable capabilities, and we are not ready for them." — Ilya Sutskever

These days, OpenAICo-founder and former chief scientist Ilya Sutskever A speech he gave sparked heated discussion.

This is his speech when he received an honorary degree from the Open University. Sutskever recounts his journey from Russia to Israel, and then to Canada, becoming...AIThe legendary stories of top researchers in the field.

Some say it's a retrospective, others say it's like a frank confession. But more accurately,It was a profound reflection on the ability to "learn"—the starting point of his life, and perhaps also...AIThe direction of the future.

Sutskever was born in Russia and immigrated to Israel with his parents when he was five years old.

He was the kind of good student who, by chance, his parents discovered the Open University, and he began to study Open University courses on his own starting in the eighth grade.

That was the first time he had it.Confidence that "can understand everything"If you read it slowly, you will definitely understand it.

Later, his family moved to Toronto. Instead of going back to high school, he went to the public library to look for...Machine LearningThe book.

That year was 2002, at that timeMachine LearningIt's still largely a niche area in academia; expert systems are the mainstream, and image recognition is still in its infancy. But he didn't follow the trend; he sought answers based on his own questions.

He said he wanted to figure out a question:Can computers learn? Is learning even possible?

This question became the central theme of his life.

Later, through a series of coincidences, he joined the University of Toronto, becoming Geoffrey Hinton's student, and together with his team created AlexNet, which opened up a new era of innovation.Deep learningThe prelude.

He and his team were acquired by Google and joined Google Brain. Later, he and several friends founded OpenWare in the Bay Area, a firm we all know.AI.

From a boy flipping through books in the library to building the world's strongestAIThe model researcher has actually been working around that problem all along:What is true learning? How is comprehension developed? Is it possible for machines to take this step?

In the middle of his speech, he talked aboutAIThe future. Everyone is looking forward to him talking about Chat.GPTRegarding the future technological landscape, he said, "We are not ready yet."

He wasn't making predictions from a stage; rather, after decades of research, he posed a question to humanity—

We understandAIDoes it understand us?

ifAIIt would be a good thing to one day be able to conduct medical research—to cure diseases, or even prolong life. But here's the problem: if...AIIf it can do all that, what else can it do?

AIHaving learned how to learn, it may then create something even smarter on its own.AI.

What should we do at that moment?

There was no answer. Neither did Ilya. He simply said, "We need to be prepared." Although there was no clear method at the moment.

These words, spoken by him, carry a different weight. He doesn't think about the future in a "CEO-like" way like Altman, nor does he sound like a warning bell like Musk. He's more like a researcher, constantly asking himself, the machine, and the world, how far can "learning" really go?

At the end of his speech, he said that for a long time, the Open University represented his entire faith in learning.

It wasn't just a piece of paper, but a symbol. That was the first time he understood that as long as you're willing to learn, you can really go very far.

That person who relied on textbooks for self-study and searched in the libraryMachine LearningThe boy who started with the introductory book now stands in the worldAIHe said that the center of the industry is a perfect closed loop.

He didn't use words like "gratitude," but rather "faith."He doesn't believe in the education system, but in learning itself.

After watching this speech, I had a very strange feeling in my heart.

NowAIThe industry is booming, models are getting bigger every day, and products are being updated weekly. But those truly at the forefront, like Ilya, don't talk about being faster or stronger; they talk about...We don't know yet..

He spoke slowly and without sentimentality, but every word he uttered was like an echo from the depths of knowledge.

He wasn't making an accusation, nor was he peddling a vision of the future. He was simply reminding us...AIThey won't wait for us to be ready before they evolve.It's learning, and we must continue learning. We can't just rely on observing, using, and adjusting parameters.

AIWhether the future can be controlled and aligned is uncertain; no one can say for sure right now. But one thing is certain:

If humans give up learning, they will truly never understand it.

The following is the full text of the speech:

I was born in Russia, and my parents immigrated to Israel when I was five. I started school and was a good student, and my parents wanted me to study in a better environment. One day, we stumbled upon the Open University, and I started taking their classes from eighth grade. It was the best experience ever.

By the way, I got my textbooks. The Open University's textbooks are very clear, and I started reading them as soon as I got them. I understood them very well. At a certain point, I remember that moment, I felt like a completely different person. I became confident that as long as I read slowly, I would definitely understand. So this helped me a lot. Also, as you know, I studied mathematics and computer science. I built a very solid foundation.

Later, my parents moved to Toronto, and the first thing I did was go to the Toronto Public Library to try and find a book about...Machine LearningRather than repeat high school, I decided to try transferring to the University of Toronto. Luckily, Jeff Hinton was also in Toronto. This was the place to go, the most forward-thinking university in the world at the time.AIMy thoughts were all there. And so I successfully transferred to the University of Toronto.

In 2002, computers could already play some video games; they could play chess and checkers, but that was about it. I remember thinking that a very powerful computer could play chess, no problem, but how did it learn? Is learning even possible? Can computers learn?

I believe that if we can answer the question "how to learn," then everything else will follow. We wrote the Alex paper during our graduate studies, and several companies expressed interest in acquiring our company. However, we didn't actually have a company at that time, so we needed to create one. Ultimately, Google acquired us. Therefore, I joined Google. [Large]Neural NetworksThe idea that anything could be done was further confirmed in our research at Google. After that, I became involved in founding Open...AIThe opportunity. I was in the Gulf region at the time. I thought to myself, how can I not try a real entrepreneurial opportunity here? Especially with so many outstanding people, so I decided to do it. We continued at OpenAIWork.

Deep learningIt's a big idea, and many people have contributed to it. I'm glad I can contribute as well.

But speaking ofAIHowever, the future is not so promising.Simple.AIIt's very exciting because it's verypowerfulRight? Once you have the ability, you can do all sorts of things. So, when...AIBecome enoughpowerfulCan you imagine?AIWhat kind of things would he do? IfAIIf we become capable enough, we will possess incredible medical technology, ifAIBeing able to conduct medical research would be amazing. We could do so much more, cure many diseases, and even extend lives. I think these are truly wonderful things, but if...AIIf it can do that, what else can it do? The ultimate answer is: it can do everything, in ways that are both unpredictable and unimaginable.

How should we respond? We must be prepared, but there's no clear solution yet. Ultimately,AIIts power will be immense.AIAble to create the next generationAIThis is what is calledintelligentAn explosion. Oh my god, what should we do then?AIThe problem is that its impact is too great, too...powerfulIt can solve all problems, it can do anything, and all of these problems currently have no answers.

Now, I would like to express my deepest gratitude to the Open University for awarding me this honorary degree. I find it very meaningful because for a considerable period of time, the Open University has represented to me the entire academic community and all those interested in learning. It's like a perfect closed loop, in a very meaningful way.