Thinking – what is It? What does it mean for humans to think? Are we about to be outdone? artificial think? A lot of people think so, but not Teppo Felin, as far as I understand him. In this episode, EconTalk host Russ Roberts welcomes Felin to discuss these questions, based on a working paper Felin co-authored with Matthias Holweg. Felin explains that the human ability to to ignore Existing data and evidence are not only our Achilles heel, but also one of our superpowers. So what are we actually? Doing when we think???
After listening to the podcast, we are curious about your opinion. Share your answers to the prompts below in the comments or use them to start your own offline conversation. Either way, we’re here for it.
1- What’s wrong with thinking of the brain as a computer? This ubiquitous metaphor has been around since the 1950s. It’s true, Felin says, that neural networks make connections, but the brain is still not really a computer; something else is going on. How should we understand the brain according to Felin?
2- Large language models can accommodate so much content. Felin says it would take people hundreds of years to pre-train their brains for the same number of words. He describes that LLMs use a stochastic process that is very good at “predicting the next step.” If LLMs contain so much content, why can’t they think like humans? What is the significance of looking back in the process of human thinking?
3- Felin argues that using AI to make rational decisions is a fool’s game, as is the idea of being able to rid AI of biases. Why are biases a feature and not a bug of human cognition? What is the relationship between beliefs and theories and thinking, and how does this distinguish human thinking from AI?
4- Felin uses the charming example of the Wright brothers and emphasizes that no venture capitalist would have given them any money. Why would AI have done that according to Felin? hindered the progress of the Wright brothers? Why, on the other hand, were they able to succeed?
5- Roberts quotes Yann LeCun; “Prediction is the essence of intelligence.” To what extent does prediction have to do with intelligence, according to Felin? What does Felin mean when he says: “…that’s something computers can’t do. They take existing data as a given, while we as humans do it through experiments new dates.” To what extent is there a meaningful distinction between digital and biological intelligence, and how could you describe this?