Kahneman on AI and Humans
A few months ago Daniel Kahneman gave an interview to Lex Fridman about Deep Learning and AI. The discussion also naturally wandered into the spheres of expertise common to Kahneman, like cognitive biases, modes of reasoning and happiness research. Here is a link to their conversation: https://www.youtube.com/watch?v=UwwBG-MbniY
It transpired fairly quickly that Kahneman did not actually believe that AI based on Neural Networks or Deep Learning as part of Neural Networks would acquire the capacity for reasoning. Moreover, AI is totally bereft of the sense of meaning, self-awareness, interpretation and causality, especially time causality, in Kahneman’s view. At the same time Kahneman emphasized that the speed of development in the field of AI was staggering. The transition from becoming unbeatable in chess to reigning in the game of go was breathtakingly swift.
Humans learn quickly, machines learn by going through myriads of examples, often being supervised at doing so. How soon would computers be able to learn like humans or learn in some other way but as effective as humans are questions that are difficult to answer now. Kahneman believes that even with Deep Learning will hit the limits of what machines can do very soon. Partly he shares views of Gary Marcus on this topic (see Marcus’ book entitled Rebooting AI: Building Artificial Intelligence We Can Trust, coauthored with Ernest Davis; Marcus and Davis write in this book the following: “When Thales investigated electricity in 600 BC, he knew he was onto something, but it would have been impossible to anticipate exactly what; we doubt he dreamt of how electricity would lead to social networking, smart watches, or Wikipedia. It would be arrogant to suppose that we could forecast where AI will be, or the impact it will have in a thousand or even five hundred years” ).
There are, however, computer scientists who sincerely maintain that the computer architecture that we deal with at present is sufficient to take us to artificial reasoning and consciousness. Ray Kurzweil, for instance, is enthusiastic about AI development prospects – too much in Kahneman’s opinion. He stresses that reasoning by itself does not give us much. There should be also the corresponding understanding of the problem. Albert Einstein has a saying attributed to him: “Any fool can know. The point is to understand”. Computers can anticipate and predict, but whether they can understand is the question answered in negative yet. There may by problems which defy solutions without prior understanding. In other words, unless a problem is properly understood a solution to it may be unreachable.
The discussion also touched upon Kahneman’s book Thinking Fast and Slow and the corresponding cognitive systems: System 1 and System 2. Kahneman pointed out that many people interpreted incorrectly System 1. It is not intuitive in its essence. It is a skill that we have to learn and it relies on a worked-out model of the world. For instance, driving a car or speaking a language. Also, System 2 has frequently been misconstrued. Its core is that it verifies the data coming through perception and the models that our mind builds about the surrounding world.
On the other hand, we may say that every time a computer starts working and cooperation with a human the latter soon becomes superficial. Gary Kasparov once promoted the idea that a computer with a human being would be more powerful than alone. As we may see now computers do not need humans to play chess in the winning way at all.
It is also worth remembering that each of us has the experiencing self and the remembering self. The real life happens to the experiencing self. However, we make decisions relying on the remembering self. Moreover the remembering self does not really have to be faithful to what has happened. It forms stories and puts events in a way which makes these stories compelling. Come to think about it, reasons are also stories we come up with to justify this or other choice we make. When we are asked to explain why we have selected this particular option we then bring about reasons which are based on shared fictions – we tell a story that would be acceptable for others.
There is, therefore, a huge discrepancy between where theoretically happiness comes from (enjoying the events we live through, having fulfilling and delightful experiences, etc.) and where it actually lies (being able to tell a wonderful story of your life to yourself and to others). Paradoxically time does not matter in the stories, events do. Time is the main currency of life. It is, however, underrated in stories.
It is telling that Kahneman said that he had never used and even seen Instagram. Therefore he could not comment on the effects of recording one’s life through leaving a trace of pictures taken continuously of the past events. Would this magnify the importance of the remembering self?
If reading the above has triggered some interesting reflections on your part please share them with us by writing to firstname.lastname@example.org