Tesla and SpaceX boss Elon Musk has been especially vocal about the potential negative implications of the development of artificial intelligence. Musk has said AI poses "vastly more risk than North Korea," that "competition for AI superiority at national level most likely cause of WW3," that "AI is a fundamental risk to the existence of human civilization," that governments "will obtain AI developed by companies at gunpoint, if necessary," and that "AI is far more dangerous than nukes."
Eric Schmidt, the former executive chairman of the Board of Directors of Alphabet, Google's parent company, has recently come down in direct opposition to Musk's dire point of view.
In December, Schmidt transitioned from executive chairman of the Board of Directors to technical advisor for Alphabet, while continuing to serve on the board, according to a statement at the time
If you're not concerned about AI safety, you should be. Vastly more risk than North Korea.
According to Schmidt, the reason Musk is "exactly wrong" is that he doesn't understand the full ramifications of the potential of artificial intelligence, he said in Paris.
"And he is wrong because he doesn't understand the benefits that this technology will provide to making every human being smarter. And the fact of the matter is that AI and machine learning are so fundamentally good for humanity," said Schmidt.
"Over and over again, making people smarter is a net good. He is concerned about the possible misuse of this technology. And I am too. But today, the overwhelming benefit of this is positive."
Schmidt continued with an example: "You would not invent the telephone because of the possible misuse of the telephone by evil people. No. You would build the telephone and you would try to figure out a way to police the misuse of the telephone," Schmidt said.
One current application of artificial intelligence is Google translate, said Schmidt. In the future, artificial intelligence will be used to power self-driving cars and to improve medical care, he said.
Currently, artificial intelligence should operate in conjunction with humans, said Schmidt.
"At the moment it has too many errors to fly the airplane. Okay? So it's fine for the AI to advise the pilot but you do not want AI flying your airplane. Too many errors, too many false positives," said Schmidt in Paris. "Now, over time, these errors will become smaller, and eventually perhaps zero."
Another problem with current artificial intelligence technology is that computers are not able to explain how they came to an answer.
"And the other problem with the systems we are talking about is in the big language and learning models, they can't exactly tell you how they learned something," says Schmidt.
"So the example is you have got a car and the car makes a wrong turn and the car is self-driving. And so the accident board says to the car, 'How did you learn this?' And the car truthfully says, 'I don't know how I learned that.' Because they learn it in a different way," says Schmidt.
"Again, there are people working on this problem. But until we have real accuracy, AI and machine learning need to be advisory. Advice to you. So I like to think of it not as a replacement of humans â?? which I don't think it will be at all â?? but rather as making you smarter."
But despite these current hurdles, Schmidt sees artificial intelligence becoming as widespread in society as computers are today.
"When I was young, computers were the regime of very specialized people. They were very hard to use, very hard to understand. Twenty years later, computers through the Macintosh and Apple and the IBM personal computer ultimately brought us to the iPhone and Android phones that you have today. So we went from very specialized very hard to understand solutions to very general purpose solutions and we will see the same thing with AI," says Schmidt.
"This technology that is today the realm of specialists will become generally useful for humans. Of that I am sure," the Google boss said in Paris.
The article was originally published here