In conversations emphasized by luckHinton explained that AI research agendas are increasingly shaped by short-term economic benefits rather than broader issues regarding human survival. “For corporate owners, it's the short-term benefits that drive research,” he said, adding that curiosity-driven research often overlooks a larger picture of what AI means for the future of humanity.
Beyond fraud: economic and existential risks
Hinton, who supported the pioneering artificial neural networks, has long warned that unidentified AI could widen wealth gaps and disrupt the labor market. Now he emphasizes that there are two dangers. On the other hand, there is an immediate risk Bad actor Use AI to design cybercrime, misinformation, and even harmful viruses. Meanwhile, there is the horrifying possibility that AI itself will evolve into a powerful actor who may not have the best interests of humanity.
These concerns will only get worse if companies focus solely on profits, he says. “This is very different from the risk that AI itself becomes a bad actor,” Hinton pointed out, emphasizing that today's incentives to drive AI development ignore the catastrophic potential of tomorrow's technology.
The mother's instincts as a fundamental solution
Earlier this year, Hinton proposed an unconventional idea at the AI4 conference. We have embedded “maternal instincts” into the AI system. Drawing from nature, he argued that the only consistent example of a more intelligent being being “controlled” by something less intelligent is the relationship between a mother and her baby. Peaceful coexistence may be possible if machines are taught to care for human happiness in a similar way.
“If AI is going to make me a parent, it's going to replace me,” warned Hinton, estimating there is at least 10% chance of an AI-driven extinction within the next 30 years. His timeline for the arrival of close AI has also been dramatically shortened. He now suggests that it can arrive in just 5-20 years.
Calling to act before it's too late
According to Hinton, urgency lies in changing resources. Today, most funds are directed at making AI stronger and less secure. Without an investment in alignment research, machinery strengthening is trained to act in the interests of humanity, but risks can spiral beyond control. As fraudsters continue to exploit AI in predictable ways, Hinton's message is that greater danger lies elsewhere. “We must decide what value we want in our ai' children before we can surpass us,” he warns. He will warn him that he will wait a long time, which means it's already too late.
