Don't let AI threaten the survival of humanity

The 'godfathers of AI', winners of the VinFuture 2024 prize, believe that scientists need to ensure that AI scenarios do not threaten the survival of humanity.

Today, December 7, the winners of the VinFuture 2024 Prize interacted with students and young scientists at VinUni University about scientific research as well as the future prospects of the fields they are pursuing. Here, the winners of the main prize, outstanding scientists whose groundbreaking works have laid the foundation for today's AI revolution, talked about the amazing advances of AI.

Professor Yann LeCun (middle) and Professor Yoshua Bengio (left cover) at the exchange session

"We feel like we've jumped 10 years into the future."

Not coming to Vietnam to receive the award, Professor Geoffrey Hinton, University of Toronto, Canada, participated in the video exchange. He, Professor Yann LeCun and Professor Yoshua Bengio, all received the Turing Award (an award considered the Nobel Prize in the field of computer science) in 2018, and are all called "AI godfathers" by the international media. Now, all three of them continue to be honored by VinFuture 2024 in the main award along with two other scientists, Professor Fei Fei Li (USA) and Mr. Jensen Huang (CEO NVIDIA).

The recent explosion in AI is the result of three factors, says Hinton. First is the work to develop more efficient ways for neural networks to learn, which he, Yoshua, Yann, and many others have done. Then there’s the incredibly fast computing power of graphics processing units (GPUs) and the vast amounts of data we can collect.

Professor Geoffrey Hinton shares via video

“A great aspect of this award (VinFuture 2024) is the recognition of Mr. Jensen Huang's work in pioneering the creation of GPU boards that are useful for AI. The combination of Jensen Huang's GPU board and Professor Fei Fei Li's datasets played an important role in the modern development of AI. It is great that their work is recognized alongside those working on neural network algorithms,” said Professor Hinton.

“We found that when we used GPUs to train neural networks, they were 30 times faster. That was a quantum leap. It was like the computing speed was 10 years ahead of its time. Suddenly we were 10 years ahead of the future in terms of computing speed. That was a complete game changer for AI, it was amazing!” said Professor Hinton.

Scientists are still working to understand more about the connections in neural networks. This will help scientists create artificial intelligence that is more like human intelligence, and can even explain intuition. Traditional logical approaches will never be able to do that.

"I'm starting to get scared about AI"

Professor Hinton said that today's large neural networks have intuition in a similar way to humans. Last year, he began to worry that it could take just five to 20 years for AI to be as smart as humans. "That's when I started to worry about AI. Stopping development is not an option. And so the only option is to try to figure out how to make sure it's safe when it becomes smarter than us," Professor Hinton shared.

“But I don’t think the end of the world is inevitable,” says Professor Hinton. “But the possibilities cannot be ignored, and we need to work hard to prevent it. Hopefully, the most talented students will choose to study AI safety, addressing a wide range of threats, from long-term threats like AI taking over to short-term threats like cybercrime.”

Professor Yoshua Bengio and Professor Yann LeCun

Professor Yann Lecun said he is also concerned about the risk that AI will dominate humans when it is as smart as humans. However, the current answer is not yet. "We can lose when playing chess with a small toy worth 30 euros. Is that dangerous? The answer is no. AI only becomes dangerous when we create motivation (to control humans) for AI. But right now, AI only has knowledge, not motivation," said Professor Yann.

According to Professor Yann, creating AI that actively serves humans is just a technical problem. Because AI is smart, thinking that AI will dominate humans is a prejudice, because reality shows that in society, there have been many smart people but they have not dominated the world! AI is still a tool, the problem is with humans, not AI.

According to Professor Yoshua Bengio, we cannot be subjective, just a small problem can affect the survival of humanity. Humans can put motivation into machines. For example, someone wants to turn the machine into a human, making the machine appear to have the need to "survive", we want to turn off the machine but it refuses to turn off. "Of course we have to make sure that scenario does not happen", Professor Bengio said.

Source: https://thanhnien.vn/vinfuture-award-2024-khong-de-ai-de-doa-su-ton-vong-cua-nhan-loai-185241207195331155.htm