Researchers from the University of California, Berkeley (USA) have proven that even the most advanced AI models, such as ChatGPT, cannot match the intelligence and creativity of children. Mail.ru reports this with reference to the journal Perspectives on Psychological Science.
Within the framework of the experiment, a group of children aged 3 to 5 years, as well as several adults, competed with OpenAI's ChatGPT chatbot in solving problems that required non-standard use of various objects.
Children and adults easily coped with the tasks, finding non-trivial but functional solutions, using objects for indirect purposes. This suggests that humans, unlike AI, are capable of abstract thinking, permitting them to see the hidden capabilities of objects.
ChatGPT demonstrated the limitations of its capabilities. He could only identify superficial connections between objects using standard associations, without possessing the human talent for finding new functions of objects.
Despite this, scientists are in no hurry to write off AI. They believe that chatbots, which can organize information and provide people with access to vast knowledge bases, can become valuable assistants in the development of new, more advanced artificial intelligence models.
