AI Can't Tell You Why It Did Something

The problem comes when the database and the engine go from )ach to oracle. It happens quite often that I will ask one of the students about a move from one of their games, and why he made it. If the move comes early on, the answer is almost always, "Because that's the nain line." That is, that's the theoretical move in the database, likely 5layed by many Grandmasters before. Sometimes the move isn't thery, but the student prepared it with the help of an engine, so the anwer is similar: "It's ...

 Chess is the Drosophila of Reasoning

Much as the Drosophila melanogaster fruit fly became a model organism for geneticists, chess became a Drosophila of reasoning. In the late 19th century, Alfred Binet hoped that understanding why certain people excelled at chess would unlock secrets of human thought. Sixty years later, Alan Turing wondered if a chess-playing machine might illuminate, in the words of Norbert Wiener, “whether this sort of ability represents an essential difference between the potentialities of the machine and ...

 New Kind of Memory for AI

AI researchers have typically tried to get around the issues posed by by Montezuma’s Revenge and Pitfall! by instructing reinforcement-learning algorithms to explore randomly at times, while adding rewards for exploration—what’s known as “intrinsic motivation.” But the Uber researchers believe this fails to capture an important aspect of human curiosity. “We hypothesize that a major weakness of current intrinsic motivation algorithms is detachment,” they write. “Wherein the a...

 How the Internet's Consensus of Information Undermines Au...

Heretofore, the technological advance that most altered the course of modern history was the invention of the printing press in the 15th century, which allowed the search for empirical knowledge to supplant liturgical doctrine, and the Age of Reason to gradually supersede the Age of Religion. Individual insight and scientific knowledge replaced faith as the principal criterion of human consciousness. Information was stored and systematized in expanding libraries. The Age of Reason originated ...

 Teacherbots

Another class of AI products that we can expect will be teaching machines, "teacherbots." These machines will adapt to the intelligence, knowledge, interest and curiosity levels of individual users. Human students will be able to learn at their own individual rates, instead of the incredibly clumsy schooling methods we use now. In today’s schools, a single human teacher attempts to educate a few dozen students simultaneously, pitching the intellectual level of the presentation at the middle...

 A good game teaches you how to play it

What you eventually discover as you continue to play is that Portal is a game about escaping from rooms that operate according to rules you are unaware of. You learn that each room is a puzzle, increasingly booby-trapped, and the game requires you to understand more and more complex physics in order to get out. If you don’t teach yourself the physics of each new room— that is, if you don’t learn the rules of the game—you’ll be stuck there forever, listening to the AI system repeat h...

 How Do You Find Emergent Intelligence on the Internet?

How would we know if there was an autonomous conscious superorganism? We would need a Turing Test for a global AI. But the Turing Test is flawed for this search because it is meant to detect human-like intelligence, and if a consciousness emerged at the scale of a global megacomputer, its intelligence would unlikely to be anything human-like. We might need to turn to SETI, the search for extraterrestrial intelligence (ETI), for guidance. By definition, it is a test for non-human intelligence...

 The Search for Internet Intelligence

While detecting an ET intelligence would overturn terrestrial religions forever, detecting a global internet intelligence would have wide-ranging ramifications for society. We'd have daily contact with an AI much larger than us, one that presumably would be steadily increasing in power every 18 months (Moore's Law). And this AI is embedded in the central nervous system of our global economy and culture. It's what we are connected to 24/7. It is also increasingly acts as our exo-brain. If it h...

 The Problem of Machine-Aggregated Knowledge

he nuts and bolts of artificial-intelligence research can often be more usefully interpreted without the concept of AI at all. For example, in 2011, IBM scientists unveiled a “question answering” machine that is designed to play the TV quiz show Jeopardy. Suppose IBM had dispensed with the theatrics, and declared it had done Google one better and come up with a new phrase-based search engine. This framing of exactly the same technology would have gained IBM’s team as much (deserved) rec...

 Public Policy Shouldn't Bet on Science

You ask whether, given a choice, I would put more resources into space or AI. My answer is that either choice would be stupid. Politicians always want to make such choices too soon, because they imagine they can pick winners. Usually they pick losers. The only way to improve the chances for finding winners is to keep all the choices open and try them all. That is particularly true for space and AI, which are not really competing with each other. They are done by different kinds of people in d...