What is the story
“AI Godfather” Jeffrey Hinton has expressed concern about the future of artificial intelligence (AI). In a recent episode of One decisionhe warned that AI could quickly develop its own language. That means that humans cannot understand. “Now, AI systems do what is called “the chain of thought” in English, so we can follow what it is doing. But it becomes even more frightening if they develop their own internal language to talk to each other,” he said.
Internal language for communication
Hinton further emphasizes that AI can already produce “bad” ideas, and it's not far from imagining a machine that ultimately thinks beyond human tracking and interpretation. The warning comes from his knowledge of how quickly these systems can learn and share information.
The delayed realization
Despite being a pioneer in the field, Hinton now admits that the recognition of the potential dangers raised by AI has been delayed. “I should have realised much earlier what the final danger would be,” he said. He believes that as digital systems become more refined, the gap between machine intelligence (ML) and human understanding grows at an astonishing rate.
Digital System v/s Human Brain
Hinton also highlighted how digital systems differ fundamentally from the human brain. He said, “If 10,000 people learn something and they all know it instantly, that's what happens in these systems.” This kind of collective intelligence will soon enable machines to surpass our understanding. AI models like GPT-4 already outperform humans with general knowledge, and gaps are closing rapidly even in complex inference tasks.
Disregarding the risk
Hinton has spoken out about the risks of AI, but he says that many in the tech industry are underestimating these risks. “A lot of people in large companies are underestimating the risk,” he said. However, he praised Google Deepmind CEO Demis Hassabis for his commitment to addressing these risks.
Necessity for ethical guidelines
As AI capabilities expand and governments rush to catch up, global debate over regulations is intensifying. The White House recently announced an AI Action Plan to speed up innovation. However, Hinton believes technological advancements should come with ethical guidelines. He argues that the only real hope lies in making AI “guaranteed and merciful.” Considering these systems, challenging work may soon work beyond our understanding.
