Episode 50: The Turing Test 2.0 (aka is LaMDA Sentient?)

Blake Lemoine, the ex-Google engineer, claims LaMDA -- Google's language model -- is sentient. Is he right? Alan Turing is perhaps most famous for his "Turing Test" which is a test of intelligence. David Deutsch has some interesting things to say about the Turing Test in "The Beginning of Infinity." Unfortunately, Deutsch's critique of the Turing Test is often misunderstood and it has led to some of his fans disparaging the Turing Test in ways that don't make sense.  The key question is why can humans so easily -- with a high degree of accuracy -- tell if they are talking to an intelligent being or not by merely having a conversation with the person? What is special about conversation that allows it to be used as a highly accurate test of general intelligence?  We also present a Turing Test 2.0 that improves upon the original Turing Test by removing the element of deception and formalizes the test better.  Along the way we answer the following questions: Is Blake Lemoine right that LaMDA is sentient? How can we know? Under what circumstances can a chatbot pass the original Turing Test 1.0? Will we ever have a chatbot that can pass the Turing Test 2.0? What can we learn from the Turing Test about intelligence?