Beware that devices are starting to think, understand and laugh at jokes

Competent scientist: "Artificial intelligence" will be smarter than humans after only 5 years

Jeffer Hinton has warned of the dangers of exciting technology weeks after leaving his job at Google.

In a recent interview with NPR, Hinton said there was "a serious risk that things will get smarter than us sometime soon, and that these things could get badly motivated and take over."

He stressed that politicians and industry leaders need to think about what needs to be done, regarding this issue right now.
"It's no longer science fiction," he said. "Technological progress is a serious problem that is likely to emerge very soon."

Hinton argues that the world may not be far from general artificial intelligence, which has the ability to understand or learn any intellectual task a human can perform.

Hinton retracted his previous belief that reaching this level would not be 30 to 50 years ago, noting that it could be much closer, perhaps only 5 years away.

Hinton said AI has been trained to comprehend, adding: "I'm not saying it's conscious. I also don't deny that."

"They can definitely think and they can definitely understand things."

He said he was "upset" by how smart Google's PaLM model was, noting that he understood the jokes and why they were funny.

With the launch of such AI, concerns about job replacement, political conflicts and the spread of disinformation are raised by AI.

While some leaders — including Elon Musk, who has his own stake in AI — have signed an open letter "to immediately stop for at least 6 months to train more powerful AI systems like GPT-4," Hinton doesn't believe it's possible to stop the research.

"The research will take place in China if it doesn't happen here," he said.

He highlighted that there will be many benefits to AI and stressed that leaders need to put a lot of resources and efforts into seeing if it is possible to "maintain control even when they are smarter than us."