Godfather Of AI Says Artificial Intelligence Could Destroy Humanity In 30 Years
Hinton went on to explain that the existential threat posed by advanced AI primarily comes down to our ability to control something that is smarter than we are. "How many examples do you know of a more intelligent thing being controlled by a less intelligent thing? There are very few examples," he stated. He made the comparison to a mother and her child; a toddler can control its mother, but that's through instincts that have evolved over millions and millions of years. We've developed alarmingly capable AIs in just a few decades of research.
The professor's concerns aren't necessarily unfounded. He's correct when he says that even current-generation AIs are, in some ways, much smarter than humans. All you need to do is ask ChatGPT a few unrelated questions to see this in action; the model can instantly produce accurate and detailed information on a huge variety of topics that would likely require a human to do at least a little research before answering.
The Guardian gives a better definition of AI in its reporting on this story: "computer systems performing tasks that typically require human intelligence." In that context, AI is certainly better at many things than the average human; if the ChatGPT example isn't convincing, try joining the Midjourney Discord server and generating a few images. Even experienced artists could take days to produce some of the images that modern AI generators can spit out in a few seconds. Now apply that kind of competence to project management, or military logistics and you start to see where Hinton's concerns arise.
However, as impressive as these capabilities are, there's not much in the way of actual reasoning going on behind the scenes in even the most advanced models. Despite what OpenAI claims, there's little evidence that we're making much progress toward true AI, or what is now referred to as "AGI" because the term "AI" has been poisoned through egregious misuse by marketing managers. AGI is of course "Artificial General Intelligence" and refers to self-aware software that possesses true reasoning abilities.
Professor Hinton's primary goal with his statements is to encourage governments to regulate the use and development artificial intelligence. He notes that the pace of AI development has far outstripped what he expected and he worries that "leaving it to the profit motive of large companies" will result in a negative outcome, possibly including the extermination of humanity. Governments move slowly, though, and the regulation he expects may not come soon—especially considering the lobbying power of tech giants.
