New Newsletter From Michael Harris

 Mathematician Michael Harris, author of "Mathematics Without Apologies," announces a new newsletter to concern itself with AI and the mechanization of mathematics:

https://mathematicswithoutapologies.wordpress.com/2021/07/21/announcing-a-new-newsletter-on-mechanizing-mathematics/

Robots Creating Robots

If you missed Terry Gross's "Fresh Air" episode yesterday on artificial intelligence, it's worth a listen here, when you get a chance:
https://www.npr.org/2018/03/15/593863645/robots-are-now-creating-new-robots-tech-reporter-says

Is Bluffing Just Another Algorithm...

"You've got to know when to hold 'em
Know when to fold 'em
Know when to walk away
Know when to run
You never count your money
When you're sittin' at the table
There'll be time enough for countin'
When the dealin's done.

Every gambler knows
That the secret to survivin'
Is knowin' what to throw away
And knowin' what to keep
'Cause every hand's a winner
And every hand's a loser
And the best that you can hope for
Is to die in your sleep."

-- Kenny Rogers, The Gambler

First computers mastered chess, then Go, and now AI is facing the complexity of everyone's favorite, Poker.

 A program called DeepStack is taking on the  “10¹⁶⁰ possible paths of play for each hand in heads-up no-limit Texas hold’em”:

https://www.technologyreview.com/s/603342/poker-is-the-latest-game-to-fold-against-artificial-intelligence/

Professional poker players have been defeated and researchers say they may indeed be on to "a significant advance in game-playing AI.” Another (20-day) tournament just got underway in Pittsburgh, testing Liberatus (another poker-playing bot) against an expert field. You can follow along more here:

https://www.riverscasino.com/pittsburgh/BrainsVsAI/

Prepare For Our Poker-playing Overlords

via WikimediaCommons

In a new Wired article, Adam Kucharski explains why poker may be more difficult/interesting to AI researchers than either chess or Go, where all strategic information is right in front of the players:

http://www.wired.co.uk/article/poker-ultimate-test-for-machine-learning

He quotes chess master Garry Kasparov (who lost to IBM's Deep Blue computer in 1997), saying that computers play games like chess and Go "like a machine." And then writes further,

"Kasparov hoped that games such as poker would be different. You cannot win by following a fixed set of rules because some cards are hidden, and your information is imperfect. The same is true of many other situations in life, from negotiations to auctions and trading."

Kucharski reports that the latest poker-playing robots "are revealing new and innovative ways of juggling risks and making decisions with imperfect information" and "The world's top poker bots have taught themselves to bluff, feign aggression and even manipulate their opponents."

One successful poker bot from Canada that Kucharski cites (and that progressively learns "by playing billions of simulated games") is "Cepheus" (specifically for a limit version of Texas hold 'em):

https://en.wikipedia.org/wiki/Cepheus_(poker_bot)

http://poker.srv.ualberta.ca

And, no doubt, more are on the way.

Hard and Easy Problems...

For today's Sunday reflection, Steven Pinker (from "The Language Instinct"):

“The main lesson of thirty-five years of AI research is that the hard problems are easy and the easy problems are hard. The mental abilities of a four-year-old that we take for granted – recognizing a face, lifting a pencil, walking across a room, answering a question – in fact solve some of the hardest engineering problems ever conceived…. As the new generation of intelligent devices appears, it will be the stock analysts and petrochemical engineers and parole board members who are in danger of being replaced by machines. The gardeners, receptionists, and cooks are secure in their jobs for decades to come.”   

Prepare To Submit...

...to our Google Masters (perhaps):

I've never played "Go" in my life, but that hasn't stopped me from finding the current Man vs. Machine story of a human grand champion vs. Google's AlphaGo program fascinating... fascinating specifically, of course, because Google's AI program has already won the first 2 matches (as of this moment), of a game far more complex than chess, and needs only win one more to take the best-of-5 series (it could all be over later today -- ADDENDUM: AlphaGo won the 3rd match to take the series); much to the shock of the loser, 18-time world champion, grandmaster Lee Sedol (not to mention 1000s observing the matchup).

First, driverless cars, now a world champion level Go player; how much longer can it be before a Google computer proves the Riemann Hypothesis? ;-)

Anyway, read up on the remarkable milestone matchup to date:

http://www.theverge.com/2016/3/10/11191184/lee-sedol-alphago-go-deepmind-google-match-2-result

http://tinyurl.com/jumz53c  (Slate)

http://www.economist.com/news/science-and-technology/21694540-win-or-lose-best-five-battle-contest-another-milestone

And more basic information about AlphaGo here:

https://en.wikipedia.org/wiki/AlphaGo

https://deepmind.com/alpha-go.html

For those unfamiliar with this ancient game, a 15-min. YouTube tutorial on the basics:

Moravec's Paradox

This isn't exactly math, but it's artificial intelligence (AI), and that's close enough... especially since a few posts back I wrote about IBM's "Deep Blue" and its 1997 defeat of chess grandmaster Gary Kasparov (at the time, a long-held goal of AI). Well, Moravec's paradox is the interesting idea that advanced or high-level reasoning and logic is much more easily mimicked by a computer system than are low-level sensori-motor skills that have evolved over millions of years... it's easier for a computer to learn to play chess, than to recognize human faces. This is one of those things that is fairly obvious when you stop to think about it... but, we often don't stop to think about it!
Here's what Steven Pinker wrote in "The Language Instinct":

“The main lesson of thirty-five years of AI research is that the hard problems are easy and the easy problems are hard. The mental abilities of a four-year-old that we take for granted – recognizing a face, lifting a pencil, walking across a room, answering a question – in fact solve some of the hardest engineering problems ever conceived…. As the new generation of intelligent devices appears, it will be the stock analysts and petrochemical engineers and parole board members who are in danger of being replaced by machines. The gardeners, receptionists, and cooks are secure in their jobs for decades to come.”   

A more recent blog piece applies the paradox to Google's self-driving cars, a creation I've certainly had trouble comprehending, given the countless issues/variables involved:

http://www.eugenewei.com/blog/2014/10/13/moravecs-paradox-and-self-driving-cars

[p.s. -- actually, where are the dang flying jetpacks I grew up believing we would all have by now... forget the cars Google, I want my personal commuting jetpack!]

anyway, below, another somewhat provocative post applying Moravec's paradox to brain processing:

http://blog.jim.com/science/moravecs-paradox-rna-and-uploads/

"The Man vs. The Machine"

(via MichaelMaggs/Wikimedia)

Math fans usually like chess, so I'll refer readers to FiveThirtyEight's first mini-documentary film (17 mins.), on the historic 1997 match between then-World-Champion Garry Kasparov and IBM's "Deep Blue" (actually it's the RE-match that Kasparov LOST). Some interesting history... and following its victory and acclaim, Deep Blue "retired":

http://fivethirtyeight.com/features/the-man-vs-the-machine-fivethirtyeight-films-signals/

ADDENDUM:  I've now discovered, for the more-thoroughly chess-ensconced (who have 90 minutes to devote to the Kasparov/Deep Blue battle), this older film on the same topic: