AI may soon speak in a language beyond human grasp
What's the story
Geoffrey Hinton, the "Godfather of AI," has raised concerns about the future of artificial intelligence (AI). In a recent episode of One Decision, he warned that AI could soon develop its own language, one that humans won't be able to understand. "Right now, AI systems do what's called 'chain of thought' reasoning in English, so we can follow what it's doing. But it gets more scary if they develop their own internal languages for talking to each other," he said.
Information
Internal languages for communication
Hinton further stressed that AI has already shown it can generate "terrible" thoughts, and it's not far-fetched to imagine machines eventually thinking in ways beyond human tracking or interpretation. The warning comes from his knowledge of how quickly these systems can learn and share information.
Reflection
A delayed realization
Despite being a pioneer in the field, Hinton now admits to a delayed realization of the potential dangers posed by AI. "I should have realized much sooner what the eventual dangers were going to be," he said. He believes that as digital systems become more sophisticated, the divide between machine intelligence (ML) and human understanding will grow at an alarming rate.
AI capabilities
Digital systems v/s the human brain
Hinton also highlighted how digital systems are fundamentally different from the human brain. He said, "Imagine if 10,000 people learned something and all of them knew it instantly, that's what happens in these systems." This kind of collective intelligence could soon enable machines to outpace our understanding. AI models like GPT-4 already surpass humans in general knowledge, with the gap closing fast even in complex reasoning tasks.
Industry response
Downplaying the risk
While Hinton has been vocal about the risks of AI, he says many in the tech industry are downplaying these dangers. "Many people in big companies are downplaying the risk," he said, despite their private concerns. However, he praised Google DeepMind CEO Demis Hassabis for his commitment to addressing these risks.
Ethical considerations
The need for ethical guidelines
As AI capabilities expand and governments rush to keep up, the global debate over regulation is heating up. The White House recently announced an "AI Action Plan" to speed up innovation. However, Hinton believes technical progress should be accompanied by ethical guidelines. He argues that the only real hope lies in making AI "guaranteed benevolent," a challenging task considering these systems may soon operate beyond our understanding.