Artificial intelligence should not be conflated with real human intelligence and it should not be depended on for crucial things like “retirement finance planning” and “medical diagnosis”, Dr Vint Cerf, the father of Internet, has told NDTV in an exclusive interview. The reason is that such systems currently are far from perfect, indicated Dr Cerf, who is currently the Chief Internet Evangelist of Google.
Artificial intelligence, he said, is a manifestation of a “more general form of machine learning” and it can reflect only what it has learnt.
Giving an example of what could be a real-life situation, he said a voice-activated AI could do “what you do not want it to do”. For example, it could open the door to a visitor but without checking who the visitor is, it could be a matter of concern.
There could be “large language models possibly doing you do not want it to do,” he said. “There are a bunch of edge cases and nuances that programmers should look into. These are the kind of things that worry me and others,” he added.
Asked if we are getting to a point where AI systems can become self-aware, Dr Cerf said, “I hope the answer is no”. Programmers, he said, need more time to understand the peculiarities of these systems. Giving an example, he said the larger AI models are often referred to as “hallucinating models” since they “conflate factual information and produce counter-factual output”.
Asked about the effect of deepfakes in the coming days, he said it could create concern in the entertainment industry. Deep fakes could be a “potential hazard” so it needs to be regulated and a mechanism should be prepared to make people creating them accountable.
