Meta scientist says AI is not as smart as dogs

OpenAI's ChatGPT is built on a large language model (LLM), meaning its AI system is trained on a huge amount of language data, allowing the chatbot to answer questions or commands posed by users.

The rapid development of AI has caused concern among technology experts. Many believe that AI could pose a danger to society if left unchecked. American billionaire Elon Musk has identified AI as one of the biggest risks to the future of human civilization.

When asked about the current limitations of AI, Yann Lecun said that generative AI trained on large language models has no understanding of the real world, because they are trained entirely on large amounts of text. Most human knowledge has nothing to do with language, so the human experience part is not "learned" by AI, LeCun said.

Current AI systems can pass the bar exam in the US but cannot load dishes into a dishwasher, something a 10-year-old can learn in 10 minutes, LeCun said.

107257338-1686828873084-gettyimages-1498441843-_r7a9820_k0cidqnk.jpeg — Mr. LeCun believes that we are still missing many important things to bring AI to the level of intelligence of dogs.

The AI scientist revealed that Meta is focusing on AI training research on video data, which is a huge challenge compared to training on LLM.

In another example of the limitations of current AI, he said a 5-month-old baby will look at a floating object and not think much of it. But a 9-month-old baby will look at the same object and be surprised because he or she realizes that the object shouldn’t be floating. We still don’t know how to replicate this ability with today’s machines. Without it, AI won’t reach the level of intelligence of humans, or dogs, or cats, LeCun said.

Professor LeCun also emphasized that in the future there will be machines that are smarter than humans, but this should not be seen as a threat. We should see it as something that benefits humans, each person will have an AI assistant to support in daily life, LeCun gave an example.

According to Mr. LeCun, works of science fiction have made us fear that if robots are smarter than humans, they will want to take over the world. However, there is no correlation between intelligence and the desire to dominate the world. AI systems are created under the control and dependence of humans.

Source link