Author Topic: 1st Chess , then Go , now US Air Force , when D&D ?  (Read 1736 times)

Offline awaken_D_M_golem

  • Epic Member
  • ****
  • Posts: 7334
  • (un-) Amazingly Unproductive
    • View Profile
1st Chess , then Go , now US Air Force , when D&D ?
« on: February 01, 2016, 05:16:30 PM »
Wired's write-up (usually positivist about)
http://www.wired.com/2016/01/googles-go-victory-is-just-a-glimpse-of-how-powerful-ai-will-be/

A real expert's review of the "GoBot" :  (warning this is very long go-specific)
https://www.youtube.com/watch?v=NHRHUHW6HQE

(edit flipped the two links)


Discuss ...

« Last Edit: July 07, 2016, 05:09:03 PM by awaken_D_M_golem »
avatar#3 , gravitational lensing edition ... I'm way on the other side of the universe but look like pretty rings

Offline bhu

  • Uncle Kittie
  • DnD Handbook Writer
  • ****
  • Posts: 14368
  • Fnord bitches
    • View Profile
Re: 1st Chess , now Go , when D&D ?
« Reply #1 on: February 04, 2016, 01:33:17 AM »
personally I look forward to the day they learn to play Halo and call us all noobs

Offline Amechra

  • Epic Member
  • ****
  • Posts: 4560
  • Thread Necromancy a specialty
    • View Profile
Re: 1st Chess , now Go , when D&D ?
« Reply #2 on: February 04, 2016, 04:12:21 AM »
Probably not within our lifetimes, to be honest.

Success in Chess and Go boils down to predicting what your opponent is going to do next. You have a limited (if really large) set of moves you have to analyse and check. Ultimately, developing a good AI for Chess or Go involves developing a more efficient algorithm for picking out "good" moves.

D&D... well, first you run into the fact that D&D has no actual victory condition, so gauging competency becomes much more difficult. Then you run into the fact that RPGs in general hinge on several things that computers are terrible at.

1) Interpreting natural language. Seriously, this is a big area of research because computers are shit at it, due to being utterly incapable of resolving ambiguities on the fly.

2) Social interaction. Oh, sure, we can pull all kinds of magic tricks to make it look like a computer is talking with you... But that's all those are.

3) Improvisation. I mean, the whole reason we have a DM and not just Friend Computer is because the DM can make value judgements for those times when the rules fall short. You'd have to code each new on-the-fly ruling and houserule into the AI every time you ran into one.

4) Bringing the Cheetos. I mean, have you ever seen a Cheetos delivery robot?  ;)

I mean, you probably could write an AI that could play a Fighter in combat pretty well? Or one that played some other relatively straight-forward character? But I can't really see one that plays any kind of character outside of combat (OK, the computer can just be that one dude that shows up just for combat, and naps through everything else).

The real problem with coding the AI would be dealing with all of the edge cases monsters introduce. Compared to developing a strategy to defeat (or flee from) an arbitrary encounter, making an AI that wins every single Go game it plays would be like making one that wins or draws every game of tic-tac-toe that it plays.

For reference, you can build a learning AI for tic-tac-toe from matchboxes and beads. Technically, you could do it for chess too, if you had enough matchboxes (you don't have enough matchboxes.)
"There is happiness for those who accept their fate, there is glory for those that defy it."

"Now that everyone's so happy, this is probably a good time to tell you I ate your parents."

Offline awaken_D_M_golem

  • Epic Member
  • ****
  • Posts: 7334
  • (un-) Amazingly Unproductive
    • View Profile
Re: 1st Chess , now Go , when D&D ?
« Reply #3 on: February 04, 2016, 05:04:54 PM »
 :) halo

I've read about some learning programs used
for de-bugging and eventually they get good at it.
Provided some of the hurdles you mention have
some sort of intro-work-around, said debugging
could be used the other way around, to locate
the nifty C.O.-ish points.

Personally, I think the d&d rule-set(s) is (are)
way bigger than the Bible and they've been
arguing about that one for a really long time.
avatar#3 , gravitational lensing edition ... I'm way on the other side of the universe but look like pretty rings

Offline FireInTheSky

  • Epic Member
  • ****
  • Posts: 3870
  • "Insight is the sudden cessation of stupidity."
    • View Profile
Re: 1st Chess , now Go , when D&D ?
« Reply #4 on: February 04, 2016, 09:30:13 PM »
(you don't have enough matchboxes.)
:lol

Offline JohnnyMayHymn

  • Hero Member
  • ***
  • Posts: 750
  • Former Lord of the Kitchen Sink
    • View Profile
    • Faux Blast
Re: 1st Chess , now Go , when D&D ?
« Reply #5 on: February 05, 2016, 12:16:08 AM »
Reminds me of this http://www.wired.co.uk/news/archive/2013-04/12/super-mario-solved

Here, the AI ends up stacking Tetris blocks as fast as possible, only to pause the game right before losing, to avoid losing.

Maybe we just don't tell the AI about the 'rocks fall, everybody dies' thing.
The Emperor
Can you find the Wumpus?

Offline awaken_D_M_golem

  • Epic Member
  • ****
  • Posts: 7334
  • (un-) Amazingly Unproductive
    • View Profile
Re: 1st Chess , now Go , when D&D ?
« Reply #6 on: February 06, 2016, 04:06:27 PM »
Heh.
That's ~like Data trying to not lose
against the world champ of wtf game
and winning that way. (Star Trek ng)

Driverless cars are way more important
than d&d ... not to me they aren't, but duh.
avatar#3 , gravitational lensing edition ... I'm way on the other side of the universe but look like pretty rings

Offline Solo

  • DnD Handbook Writer
  • ****
  • Posts: 1769
  • Sorcelator Supreme
    • View Profile
    • Solo's Compiled Works
Re: 1st Chess , now Go , when D&D ?
« Reply #7 on: February 06, 2016, 06:09:32 PM »
Heh.
That's ~like Data trying to not lose
against the world champ of wtf game
and winning that way. (Star Trek ng)
Strategema. And it was third level grandmaster Sirna Kolrami, thank you very much.
"I am the Black Mage! I cast the spells that makes the peoples fall down."

Offline awaken_D_M_golem

  • Epic Member
  • ****
  • Posts: 7334
  • (un-) Amazingly Unproductive
    • View Profile
Re: 1st Chess , now Go , when D&D ?
« Reply #8 on: February 11, 2016, 05:33:13 PM »
... enjoyed that episode, still memorable.
Rope-a-dope is a workable strategy (sometimes).


Computer beats race car driver :
http://www.telegraph.co.uk/news/science/science-news/11410261/Driverless-car-beats-racing-driver-for-first-time.html

and some minor Legal issue about driverless cars got cleared up within the last few days.
avatar#3 , gravitational lensing edition ... I'm way on the other side of the universe but look like pretty rings

Offline awaken_D_M_golem

  • Epic Member
  • ****
  • Posts: 7334
  • (un-) Amazingly Unproductive
    • View Profile
Re: 1st Chess , now Go , when D&D ?
« Reply #9 on: February 19, 2016, 03:45:40 PM »
Oh here we go ... Billion years+ long data storage.
Aliens are gonna be so confused by my kitty avatar  :pout
http://www.orc.soton.ac.uk/962.html

... and hey shout-out to the aliens that did find this 10 billion years in the future.  How's it hangin' guys?
avatar#3 , gravitational lensing edition ... I'm way on the other side of the universe but look like pretty rings

Offline PlzBreakMyCampaign

  • DnD Handbook Writer
  • ****
  • Posts: 1962
  • Immune to Critical Hits as a Fairness Elemental
    • View Profile
Re: 1st Chess , now Go , when D&D ?
« Reply #10 on: March 13, 2016, 02:40:04 PM »
It's just an evolution model, really. It's sort of like splicing videos together of other people doing well at something. Then when you line up the video so the motions don't look edited, you say "Look how well I continuously score!" Well, yeah.

Don't get me wrong. It is an achievement, and one that took a lot of work. But its not something inherently special. It's not real AI

personally I look forward to the day they learn to play Halo and call us all noobs
Clearly you've never played against max perfect dark or UT bots. It's simple: maphack, locate closest sighted opponent, run straight there while aiming, aim to anything gaining sight, shoot first.

Seriously. It's painful. You have to play in slow-motion to have any hope of fighting back. It's also kind of beautiful to see the laser efficiency. Now, coding the above without the maphack is actually challenging but no longer brutal in efficiency.

Aliens are gonna be so confused by my kitty avatar  :pout
... and hey shout-out to the aliens that did find this 10 billion years in the future.
We don't understand that kitty, and we're people. But nice acknowledgement of the time-depth before any kind of intelligent interstellar travel would make it here. That's far more realistic than zooming by casually for some butt-probing, cow slaughter, and grass patterns.

@Amechra with enough code, you could make a program that played DnD as a simple PC. What you couldn't do is make a program that DMd. That's the difference between making routines for a new DotA hero's AI and asking a programmer to make a program that programs sufficiently interesting new dota mechanics (heroes, items, terrain features, etc). The first is doable, if incredibly tedious. Yes, it won't be perfect, but it will usually perform decently well. 2) is up in the air and 3 is right out. You could only sub a computer bot for 1 player. You couldn't, say, run a whole party of them and expect anything beyond simple combat.

Offline awaken_D_M_golem

  • Epic Member
  • ****
  • Posts: 7334
  • (un-) Amazingly Unproductive
    • View Profile
Re: 1st Chess , now Go , when D&D ?
« Reply #11 on: March 13, 2016, 03:11:06 PM »
by-the-way, Lee Sedol the Go champ lost 3 straight before getting a resignation from AlphaGo.
Amazing.


Aliens ... yeah I have no idea, let's say maybe 1 per Billion years, and we wouldn't see anything except for broken down tech.  And similar for whatever comes after.  But oh what fun you'll have in an undistinguishable-from-our-reality simulation with my kitty avatar's Tail  playing the Dr.Suess + Cthulhu combo.  Oh what fffphl-fun !!
avatar#3 , gravitational lensing edition ... I'm way on the other side of the universe but look like pretty rings

Offline JohnnyMayHymn

  • Hero Member
  • ***
  • Posts: 750
  • Former Lord of the Kitchen Sink
    • View Profile
    • Faux Blast
Re: 1st Chess , now Go , when D&D ?
« Reply #12 on: March 24, 2016, 03:24:16 AM »
maybe if 'D&D' is being used loosely, we could make a game like D&D which is easier for an AI to DM
The Emperor
Can you find the Wumpus?

Offline awaken_D_M_golem

  • Epic Member
  • ****
  • Posts: 7334
  • (un-) Amazingly Unproductive
    • View Profile
Re: 1st Chess , now Go , when D&D ?
« Reply #13 on: March 25, 2016, 04:31:05 PM »
DM-ing is hard, I won't guess as to when.

I know one of the AI's solved Poker, but the program is way too complicated for a Human to try to play.


Monopoly could be their next real target.
The #'s have been known for quite a while.

Programs couldn't handle the social side of bluffing,
and wheeling/dealing with the other players, but
one of the AI Psych-bots might figure some of this out.
What's the magic word :  heuristically? (idk)

The really fancy maths of the mortgages and
the official rules that most people don't actually follow,
are right up Wall Street's alley.  They could probably
get this part long before the Psych-bots work.

 :??? :D
avatar#3 , gravitational lensing edition ... I'm way on the other side of the universe but look like pretty rings

Offline awaken_D_M_golem

  • Epic Member
  • ****
  • Posts: 7334
  • (un-) Amazingly Unproductive
    • View Profile
Re: 1st Chess , now Go , when D&D ?
« Reply #14 on: April 29, 2016, 04:00:20 PM »
 :D
Grandma ... versus ... Driver-less Tesla

(2 minutes)
https://www.youtube.com/watch?v=MlF2d9E0gB4
avatar#3 , gravitational lensing edition ... I'm way on the other side of the universe but look like pretty rings

Offline Jackinthegreen

  • DnD Handbook Writer
  • ****
  • Posts: 6153
  • I like green.
    • View Profile
Re: 1st Chess , now Go , when D&D ?
« Reply #15 on: May 17, 2016, 10:05:09 AM »
Getting AI to optimize D&D would take a significant amount of rewriting the rules into something a computer can understand and process efficiently.  I'm somewhat looking forward to that possibility because then it might be something rules creators can look at to make for less ambiguous rules that players don't get into petty squabbles over.

Of note is computers are already getting proficient at analyzing legal documents for various discrepancies and other such things.

Offline awaken_D_M_golem

  • Epic Member
  • ****
  • Posts: 7334
  • (un-) Amazingly Unproductive
    • View Profile
Re: 1st Chess , now Go , when D&D ?
« Reply #16 on: May 17, 2016, 06:22:29 PM »
Niiiice.
But it feels rather spirit-of-the-law.
Patent Trolls are doing the same thing, but exclusively for the money.

Apply both ideas to the horrendously convoluted legal situation(s) in India.
avatar#3 , gravitational lensing edition ... I'm way on the other side of the universe but look like pretty rings

Offline awaken_D_M_golem

  • Epic Member
  • ****
  • Posts: 7334
  • (un-) Amazingly Unproductive
    • View Profile
Re: 1st Chess , now Go , when D&D ?
« Reply #17 on: July 07, 2016, 05:08:07 PM »
wtf  :???


AI beats top U.S. Air Force tactical air combat experts in combat simulation

source article ---> http://www.omicsgroup.org/journals/fuzzy-logic-based-intelligent-agents-for-unmanned-combat-aerialvehicle-control-2167-0374-1000139.php?aid=67461
~ lite version ---> http://www.kurzweilai.net/ai-beats-top-u-s-air-force-tactical-air-combat-experts-in-combat-simulation

Hey I want one of those $35 compooters, now !!
avatar#3 , gravitational lensing edition ... I'm way on the other side of the universe but look like pretty rings

Offline altpersona

  • Legendary Member
  • ****
  • Posts: 1856
  • #78
    • View Profile
    • You are here
Re: 1st Chess , then Go , now US Air Force , when D&D ?
« Reply #18 on: July 07, 2016, 05:21:51 PM »
Quote
According to IBM, the cloud-delivered enterprise-ready Watson has seen its speed increase 24 times over—a 2,300 percent improvement in performance, and its physical size shrank by 90 percent—from the size of a master bedroom to three stacked pizza boxes.


that was two years ago. next year is should half that...

with luck my grandkids will have a bot as the wage earner for the family.
The goal of power is power. - 1984
We are not descended from fearful men. - Murrow
The Final Countdown is now stuck in your head.

Anim-manga still sux.

Offline awaken_D_M_golem

  • Epic Member
  • ****
  • Posts: 7334
  • (un-) Amazingly Unproductive
    • View Profile
Re: 1st Chess , then Go , now US Air Force , when D&D ?
« Reply #19 on: July 08, 2016, 04:18:21 PM »
Heh don't be that optimistic.
There are jobs that are not Robot-able.
But lots of jobs will fall by the way side soon.

There isn't equal distibution of computing power
and people who have it and can use it productively
are obviously better off.


I mean this is an off-the-cuff old school grumble, but
most phone calls and texts, have no economic value.

I went in to get my final paycheck from Pizza Dorks.
They were having their record busiest hour.
I hopped on the phones for 1 hour. 
Manager couldn't figure how to pay me since I'd already quit.
Decided to give me $35 free food.
I knew that usually they did worker food at 1/2 off.
It was still about 5 times minimum (at the time I though awesome!).
I took about 70 calls, each one about $10 of food = $700 of business.
My labor rate was barely above 2%.
I much later found out, that high teens % was normal.
5 * 5 = 25
I worked at 1/25th the wage I deserved.  Or worse.
avatar#3 , gravitational lensing edition ... I'm way on the other side of the universe but look like pretty rings