Former Google CEO: AI is like nuclear weapons. Big countries need to establish 'AI deterrence' to ensure mutual destruction.

Former Google CEO Eric Schmidt compared artificial intelligence to nuclear weapons in an interview and called for a similar deterrent mechanism of mutually assured destruction to prevent the world’s most Powerful countries are destroying each other because of AI.

Mutually Assured Destruction (Mutually Assured Destruction, referred to as the M.A.D. mechanism, also known as the principle of mutual destruction) is an idea of ​​"all destroyed" nature. If one of the two opposing parties uses it fully, both parties will be destroyed, which is called the "balance of terror."

Schmidt compared AI to nuclear weapons, saying that China and the United States may conclude a treaty similar to the ban on "nuclear testing" in the future to prevent AI from destroying the world.

Schmidt: I was still very young and naive at the time

On July 22, Schmidt spoke at the Aspen Security Forum on national security and artificial intelligence. The dangers of artificial intelligence were discussed at the panel.

In response to a question about ethical values ​​in technology, Schmidt explained that he himself was naive about the power of information in the early days of Google.

He then called for technology to better align with the morals and ethics of those it serves, and made a bizarre comparison between artificial intelligence and nuclear weapons.

Schmidt envisions that in the near future, China and the United States will need to sign some treaties around artificial intelligence.

Schmidt said: "In the 1950s and 1960s, we ended up with a 'to be expected' rule about nuclear testing, and ultimately nuclear testing was banned."

Schmidt believes that "this is an example of the balance of trust or lack of trust, and it is a 'no surprises' rule." He is very worried about the beginning of some misunderstandings and misunderstandings between the United States and China, the two artificial intelligence powers. Something that leads to triggering a dangerous event.

Schmidt said that no one is currently doing research in this area, but artificial intelligence is so powerful.

Eric Schmidt served as CEO of Google from 2001 to 2011, executive chairman of Google from 2011 to 2015, and executive chairman of Alphabet from 2015 to 2017. From 2020 to 2020, he served as a technical advisor to Alphabet.

In 2008, while serving as chairman of Google, Schmidt campaigned for Barack Obama and later worked with Eric Lander Lander and became a member of President Obama's Council of Advisors on Science and Technology.

From 2019 to 2021, Schmidt co-chaired the National Security Commission on AI with Robert O. Work.

#Is AI really that dangerous?

Artificial intelligence and machine learning are impressive yet often misunderstood technologies. For the most part, it's not as smart as people think it is.

It can produce masterpiece-level artwork, beat humans in "StarCraft 2", and can make basic phone calls for users. However, trying to make it complete more complex tasks, such as autonomous driving, But it didn't go well.

Schmidt envisions that in the near future, both China and the United States will be concerned about security issues, forcing both sides to reach some kind of deterrence treaty on artificial intelligence. He spoke of the 1950s and 1960s, when countries used diplomacy to orchestrate a series of controls over the deadliest weapons on earth. But it would take decades of nuclear explosions for the world to get to the point where the Nuclear Test Ban Treaty, as well as SALT II and other landmark legislation, were enacted , such as the nuclear explosions in Hiroshima and Nagasaki.

The United States used nuclear weapons to destroy two Japanese cities at the end of World War II, killing thousands of people and proving to the world the eternal horror of nuclear weapons.

Subsequently, the Soviet Union and China also successfully developed nuclear weapons, and then came the creation of Mutually Assured Destruction (MAD), a deterrence theory that maintains a "balance of dangers" to ensure that if one country launches a nuclear weapon, other countries may also launch nuclear weapons. emission.

To date, humanity has refrained from using the most destructive weapons on the planet because doing so could destroy civilization across the globe.

Does artificial intelligence currently have such power? ​

It seems that artificial intelligence has not proven itself to be as destructive as nuclear weapons, but many people in power are afraid of this new technology, and people have even suggested giving control of nuclear weapons to artificial intelligence. These people believe that Artificial intelligence is better suited than humans to serve as arbiters of nuclear weapons use.

So the problem with AI may not be that it has the potential world-destroying power of nuclear weapons, AI is only as good as its designers and they reflect the values ​​of their creators.

Artificial intelligence has the classic "garbage in, garbage out" problem. Racist algorithms give birth to racist robots, and artificial intelligence will also produce biases.

DeepMind CEO Demis Hassabis understands this better than Schmidt.

DeepMind has developed an artificial intelligence capable of defeating Starcraft II players. In a July interview on the Lex Fridman Podcast, Fridman asked Hassabis how to control an artificial intelligence like Such powerful technology, and how Hassabis himself avoided being corrupted by such power.

Hassabis’s answer: “Artificial intelligence is too big an idea,” he said. “What matters is who created (artificial intelligence), what culture they come from, what values ​​they have, They are the builders of artificial intelligence systems. The artificial intelligence system will learn on its own... But the culture of the system and the values ​​of the creator will remain in the system."

Artificial intelligence is one of the The creator reflects that a 1.2-megaton explosion cannot level a city to the ground unless humans teach it to do so. ​

The above is the detailed content of Former Google CEO: AI is like nuclear weapons. Big countries need to establish 'AI deterrence' to ensure mutual destruction.. For more information, please follow other related articles on the PHP Chinese website!