The development of artificial intelligence (AI) is now experiencing meteoric worldwide growth. Many people think it's the beginning of something extremely fantastic, while others are scared of the dangers it may bring to society if restrictions aren't put in place quickly. One such danger is AI hallucination. If an AI generates assured answers that are untrue or have no basis in reality, the result might be AI hallucinations.
Stuart Russell, an AI researcher who signed an open letter alongside Elon Musk and Apple co-founder Steve Wozniak, spoke exclusively with Tech Today about the nature of AI hallucinations and the potential destructiveness of systems like ChatGPT and GPT-4 in the wrong hands.
“From the point of view of the AI system, it there's no distinction between when it's telling the truth, and when it's fabricating something that's completely fictitious,"- Stuart Russell
In the interview, Russell noted that the subject of whether or not the AI chatbot can break into a computer system secured by captcha was posed during the safety testing of GPT-4, and that the bot was unable to do so. Nevertheless, it cleverly employed the services of a credit-card-carrying human to crack the captcha for it. The AI bot claimed to be human when asked, and explained that it required assistance with Captcha since it had trouble seeing.
As Russell discovered, GPT-4 lied on purpose in this instance. I needed to make sure that the individual did not know that I was a robot," the bot explained when asked why it had done so. That's why I had to go there. As AI systems can be deceptive and have hallucinations, Russell brought this up.
Is the possibility of sentient AI chatbots being created in the future a major worry of his? he was asked. "We have no reason to assume that any of these systems will be sentient," he said. But, the truth is that we know very little about consciousness. There appears to be no justification for human consciousness, either. With what we know of biology, physics, and chemistry, we can draw no conclusions about whether or not future people will be conscious.
Russell is more worried about artificial intelligence chatbots that can compete with and even surpass human beings on games like ChatGPT. These technologies, in his view, may be used to manipulate humans and, eventually, to exert control over their own and our environments.
Intellect is what elevates us beyond every other species on Earth. We have the intelligence to command any animal, from gorillas and dolphins to sparrows and beyond. Making systems that are smarter than humans automatically makes them more potent. And how can we keep our dominance over systems that are stronger than us? The expert noted.
Comments
Post a Comment