The “sponsor of AI” reveals the only way in which humanity can survive an AI supentinant

Las Vegas
–

Geoffrey Hinton, known as the “sponsor of AI”, fears that the technology he has to build could eliminate humanity – and “technology bros” adopt the bad approach to stop him.

Hinton, a computer -winning computer scientist and a former Google executive, warned in the past that there are 10% at 20% that IA eliminates humans. On Tuesday, he expressed doubts about how technological companies are trying to guarantee that humans remain “dominant” on “submissive” AI systems.

“It will not work. They are going to be much smarter than we do. They will have all kinds of ways of getting around this,” Hinton told AI4, an industry conference in Las Vegas.

In the future, Hinton warned, AI systems may be able to control humans just as easily as an adult can weld 3 -year -old candies. This year has already seen examples of AI systems wishing to deceive, cheat and fly to achieve their objectives. For example, to avoid being replaced, an AI model tried to make an engineer sing on an affair that she learned in an email.

Instead of forcing AI to submit to humans, Hinton Presented an intriguing solution: building “maternal instincts” in AI models, so “they really care about people” even once technology becomes more powerful and more intelligent than humans.

AI systems “will very quickly develop two sub-objectives, if they are intelligent: one must remain alive … (and) the other undersender is to obtain more control,” said Hinton. “There are good reasons to believe that any type of agentic AI will try to stay alive.”

This is why it is important to promote a feeling of compassion for people, said Hinton. During the conference, he noted that mothers have instincts and social pressure to take care of their babies.

“The right model is the only model we have of a more intelligent thing controlled by a less intelligent thing, which is a mother controlled by her baby,” said Hinton.

Hinton said that it was not clear to him exactly how it can be done technically, but stressed that his critical researchers work there.

“It’s the only good result. If it’s not going to adorn me, it will replace me,” he said. “These mothers of super intelligent caregivers, most of them will not want to get rid of maternal instinct because they do not want us to die.”

Hinton is known for his pioneer work on neurons networks, which helped open the way to today’s arrow. In 2023, he left Google and started talking about the dangers of AI.

Emmett Shear, who was briefly in the acting CEO of the owner of Chatgpt Openai, said it was not surprised that certain AI systems tried to make humans sing or bypassing stop orders.

“It keeps happening. This will not stop happening,” said Shear, CEO of the Alignment startup Ai Softmax, at the AI4 conference. “The AIS today are relatively low, but they become stronger very quickly.”

Shear said that rather than trying to instill human values in AI systems, a more intelligent approach would be to forge collaborative relationships between humans and AI.

Many experts believe that the AIS will achieve superintelligence, also known as the general artificial intelligence, or acted, in the years to come.

Hinton said he thought it could take 30 years at 50 to reach act, but that this moment now sees earlier.

“A reasonable bet is between five and 20 years old,” he said.

Although Hinton remains concerns about what could go wrong with AI, he hopes that technology will open the way for medical breakthroughs.

“We will see new radical drugs. We are going to get much better treatment against cancer than the present, “he said. For example, he said that AI will help doctors to cross and correlate the large amounts of data produced by MRI and computed tomography.

However, Hinton does not think that AI helps humans achieve immortality.

“I don’t think we will live forever,” said Hinton. “I think that living forever would be a big mistake. Do you want the world to work by 200-year-old white men? ”

When asked if there was something that he would have done differently in his career if he knew how speed would accelerate, Hinton said he regretted only to focus on the work of the AI.

“I would also like to think about security problems,” he said.