Artificial Intelligence Can Learn by Thinking Like Humans

Artificial intelligence, like humans, is also capable of self-correcting and reaching new conclusions through “learning by thinking.” This is revealed by research led by Tania Lombrozo, professor of psychology and co-director of the Natural and Artificial Minds initiative at Princeton University, reported in the journal Trends in Cognitive Sciences. “There are some recent demonstrations of what looks like learning by thinking in AI, particularly in large language models,” said Tania Lombrozo.

“ChatGPT sometimes corrects itself without being explicitly told; this is similar to what happens when people are engaged in thinking learning,” Lombrozo continued. The study identified four examples of thinking learning in humans and AI: Learners can acquire new information without external input through explanations, simulations, analogies, and reasoning. In humans, explaining to a child how a microwave works can reveal our gaps in understanding.

Rearranging living room furniture often involves creating a mental image to simulate different arrangements before making any physical changes. Downloading pirated software may initially seem morally acceptable, until you make an analogy with stealing physical goods. If you know that a friend’s birthday is a leap day and that tomorrow is a leap day, you may think that the friend’s birthday is tomorrow.

Artificial intelligence exhibits similar learning processes. When asked to elaborate on a complex topic, AI can correct or refine its initial response based on the explanation provided. The gaming industry uses simulation engines to approximate real-world outcomes, and models can use simulation results as input for learning.

Asking a language model to draw analogies can lead it to answer questions more accurately than it would with simple questions. If you ask the AI ​​to think step by step, you can get answers that you wouldn’t get with a direct question. “This raises the question of why both natural and artificial minds have these characteristics,” Lombrozo said.

“I argue that learning by thinking is a kind of ‘learning on demand,’” Lombrozo added. When you learn something new, you don’t know how the information might be useful in the future. Lombrozo argues that people can put knowledge aside for later, until the context makes it relevant and it’s worth the cognitive effort to think and learn. Lombrozo acknowledges the difficulty in defining the boundaries between reasoning, learning, and other high-level cognitive functions, an area of ​​debate in cognitive science. The review also raises other questions, some of which Lombrozo intends to explore further, such as whether AI systems are actually “thinking” or simply mimicking the outputs of such processes.

“AI has gotten to the point where it is so sophisticated in some ways, but limited in others, that we have an opportunity to study the similarities and differences between human and AI,” Lombrozo said. “We can learn important things about human cognition through AI and improve AI by comparing it to natural minds,” Lombrozo continued. “It’s a crucial moment where we are in this new position to ask these interesting comparative questions,” Lombrozo concluded. (AGI)Sci/Adv