One scientist who wrote the most important textbook on artificial intelligence claims that experts are scared by their own successes in the field. He compared the advancement of AI to the creation of the atom bomb.
Professor Stuart Russell, founder of the Center for Human-Compatible AI Intelligence at University of California, Berkeley said that most experts believe that machines intelligenter than humans will be developed in the next century. He called for international treaties to regulate technology development.
According to him, the AI community is still not adjusting to the fact that they are having a real impact on the real world. This was not the case throughout the history of AI. We were mostly in the lab developing and trying to get things to work. The question of real-world impacts was not relevant at all. We must catch up quickly.
Artificial intelligence is a key component of modern life. Recent developments include machine translation and image recognition.
Professor Stuart Russell. Photograph: Peg Skorpinski
Russell, who co-authored the landmark 1995 book Artificial Intelligence. A Modern Approach and will present this year's BBC Reith lecture entitled Living with Artificial Intelligence. He says that urgent work is required to ensure humans are in control of superintelligent AI as it is developed.
AI was designed using a particular approach and methodology. He said that they were not careful enough about using this type of system in complex real-world situations.
It could prove dangerous to ask AI to quickly cure cancer. Russell said that the AI would likely find ways to induce tumours in all people, and could thus run millions of experiments simultaneously, with all of us being guinea-pigs. That's the answer to our objective. We just forgot to mention that humans can't be used as guinea-pigs, and that the entire GDP of the world can't be used to fund your experiments. You can't do this and that.
Russell stated that there was still a gap between today's AI and the AI depicted in movies like Ex Machina. However, a future where machines are smarter than humans was possible.
Russell said that I believe numbers can range from 10 years to the most optimistic to several hundred years. However, almost all AI researchers believe it will happen within the next century.
One concern is that machines would not have to be smarter than humans to pose a serious threat. He said that it is happening now. You can see that social media algorithms have enormous control over cognitive input.
He explained that algorithms are used to brainwash users, making their behavior more predictable in choosing which content to engage with. This increases click-based revenue.
Are AI researchers becoming spooked at their own successes? Russell agreed.
It reminds me of the time in physics when physicists understood that atomic energy existed. They could measure the masses and calculate how much energy was possible if they could convert between different types of elements. However, experts have always maintained that this idea was theoretical. Then it happened, and they weren't prepared for it.
He expressed concern about the use of AI in military applications, such as small antipersonnel weaponry. These are the most easily scaleable, which means you could pack a million of them into one truck, and then you could open the back to let them go. They could wipe out entire cities, Russell said.
Russell believes that the future of AI lies in creating machines that can recognize the truth and our preferences. This means they will need to consult humans before making any decisions. The idea is complicated because people have different preferences and sometimes conflicting ones.
Russell demanded that there be a code for researchers, legislation, and treaties to protect AI systems from misuse, as well as training for researchers to prevent racial bias. Russell called for the adoption of EU legislation to ban machines impersonating humans.
Russell stated that he hopes the Reith lectures will emphasize the fact that we have a choice as to what the future holds. He said that it is crucial for the public to participate in these choices because the public will ultimately benefit from them.
There was also another message. He said that progress in AI will take some time, but it does not make it science fiction.