Upon reaching a Certain Age, your teeth begin to disintegrate -- something to look forward to, kids! Thus it was that I found myself sitting my dentist's office last week, waiting for a new tooth. What I hadn't realized is that dentistry has, over the past decade, entered the era of Star Trek.
Monday, June 26, 2017
Thursday, June 22, 2017
If You Miss Next Month's Eclipse...
In just one more month, the city of Nashville will host part of the Great American Eclipse. I can already feel the excitement building, along with the email inquiries about housing in the area. If you're planning to drive somewhere to view the eclipse, book your hotel early and drive to your viewing destination well in advance. The traffic leading up to the eclipse is going to look like a hurricane evacuation in reverse.
But what if you miss the eclipse? Suppose it's overcast that day, or your car breaks down on the way? No need to despair -- there's another eclipse across the middle of the US only seven years later, in April of 2024.
But what if you miss the eclipse? Suppose it's overcast that day, or your car breaks down on the way? No need to despair -- there's another eclipse across the middle of the US only seven years later, in April of 2024.
Thursday, June 15, 2017
The Next AI Frontier: Teaching Computers to Lie
I got into a minor kerfuffle over at Steve Hsu's blog when I commented on this post about AlphaGo, the neural net program that has blown away all of the human competition in the game of Go. I said that I would be more impressed when computers mastered Diplomacy, a comment which was immediately challenged by someone else. But as impressed as I am with AlphaGo, I stand by that comment.
When I was a kid back in the 60s, the local science museum had a computer that played tic-tac-toe. (The local fair had a pigeon that played tic-tac-toe, but maybe that's a topic for another day). I have no idea how the computer worked -- maybe an analog circuit of some sort? The London science museum at one point had a tic-tac-toe computer constructed from Tinkertoys and string. As primitive as these sound, I think you can draw a straight (but very steep!) line from tic-tac-toe to the computer that "solved" checkers, and then to IBM's Deep Blue (which conquered chess), and finally to AlphaGo. These games (tic-tac-toe, checkers, chess, Go) are all deterministic with perfect information, in which the players alternate taking turns, choosing from a finite number of moves. (To be fair, there is a qualitative leap between the earlier "brute force" programs, which relied on simply increasing the number of moves scanned by the computer, and AlphaGo, in which the computer actually "learns" to play better, and in which it's impossible for the programmers to determine why the computer chose a particular move.)
Diplomacy is a very different animal. For those of you not familiar with the game, it's a contest of almost pure negotiation, requiring the ability to form alliances, offer bribes, bluff, lie, and backstab your way to the top. It requires a very different, more "human" set of skills than chess or Go. Computer scientists are already working in that direction, particularly with the game of poker.
I myself wrote two chess programs back in the 1970s, one in high school (in BASIC), and the second in college (in Fortran, the One True Programming Language). Both of my programs were, in one sense, more advanced than Deep Blue. They cheated.
When I was a kid back in the 60s, the local science museum had a computer that played tic-tac-toe. (The local fair had a pigeon that played tic-tac-toe, but maybe that's a topic for another day). I have no idea how the computer worked -- maybe an analog circuit of some sort? The London science museum at one point had a tic-tac-toe computer constructed from Tinkertoys and string. As primitive as these sound, I think you can draw a straight (but very steep!) line from tic-tac-toe to the computer that "solved" checkers, and then to IBM's Deep Blue (which conquered chess), and finally to AlphaGo. These games (tic-tac-toe, checkers, chess, Go) are all deterministic with perfect information, in which the players alternate taking turns, choosing from a finite number of moves. (To be fair, there is a qualitative leap between the earlier "brute force" programs, which relied on simply increasing the number of moves scanned by the computer, and AlphaGo, in which the computer actually "learns" to play better, and in which it's impossible for the programmers to determine why the computer chose a particular move.)
Diplomacy is a very different animal. For those of you not familiar with the game, it's a contest of almost pure negotiation, requiring the ability to form alliances, offer bribes, bluff, lie, and backstab your way to the top. It requires a very different, more "human" set of skills than chess or Go. Computer scientists are already working in that direction, particularly with the game of poker.
I myself wrote two chess programs back in the 1970s, one in high school (in BASIC), and the second in college (in Fortran, the One True Programming Language). Both of my programs were, in one sense, more advanced than Deep Blue. They cheated.
Friday, June 9, 2017
Fr. Georges Lemaitre and Cosmology
If you'd like to see my talk on Fr. Georges Lemaitre and his contributions to cosmology, there is a video posted here. You'll find me about half-way down on the right-hand side. The other talks were also very interesting and definitely worth your attention.
Thursday, June 1, 2017
The Smartest Person Who Ever Lived
Who was the greatest genius in human history? Einstein? Newton? Mark Zuckerberg? (Hint: not Mark Zuckerberg). My vote goes to someone who probably died about 3000 years ago on the shores of the Eastern Mediterranean.
Subscribe to:
Posts (Atom)