Libratus (former poker bot) now being used for military research - Poker News

Two Plus Two Forums Poker News & Discussion News, Views, and Gossip

Libratus (former poker bot) now being used for military research

Post Reply Subscribe

...

Page 1 of 2

1 2

Page 1 of 2

1 2

01-21-2019 , 03:47 AM

CyberMuppet

grinder

Join Date: Oct 2007 Posts: 581

From Wired:

"In 2017, a poker bot called Libratus made headlines when it roundly defeated four top human players at no-limit Texas Hold ‘Em. Now, Libratus’ technology is being adapted to take on opponents of a different kind—in service of the US military."

Turns out the Pentagon has funded a program to explore how the technology can be applied to military decisionmaking, and using Libratus is part of that.

https://www.wired.com/story/poker-pl...s-to-pentagon/

Quote

01-21-2019 , 09:50 AM

Xenoblade

Pooh-Bah

Join Date: Sep 2017 Posts: 4,084

lol GTO warfare

Quote

01-21-2019 , 11:25 AM

Ten5x

journeyman

Join Date: Sep 2016 Posts: 359

Quote

01-21-2019 , 11:44 AM

jukofyork

Carpal \'Tunnel

Join Date: Sep 2004 Posts: 11,749

Quote:

Originally Posted by Ten5x

Quote

01-22-2019 , 01:07 AM

zoogenhiem

old hand

Join Date: May 2016 Posts: 1,632

Joshua could play poker. It shows up in the montage too with oversized cards.

Quote

01-22-2019 , 02:07 AM

rakemeplz

Carpal \'Tunnel

Join Date: Feb 2006 Posts: 12,648

George W. Bush never knew about tight agressive play, I tell you what.

Quote

01-22-2019 , 02:21 AM

fishfood69er

old hand

Join Date: Feb 2014 Posts: 1,702

Quote:

Originally Posted by Xenoblade

lol GTO warfare

"now shoot our own soldiers 46 percent and the enemies soldiers 53 percent of the time, we must remain balanced here. others wise were too exploitable. "
-libratus

Last edited by fishfood69er; 01-22-2019 at 02:27 AM.

Quote

01-22-2019 , 03:54 AM

gamboneee

Carpal \'Tunnel

Join Date: Jun 2010 Posts: 9,163

didnt this get mentioned a week or two ago?

stalin would be king of that style gto war

Quote

01-22-2019 , 04:57 PM

PTLou

JellyBean

Join Date: Nov 2008 Posts: 5,516

in all seriousness, could someone that understands Libratus and GTO please answer this.

I dont understand the "Strategy Robot" introduced but not clearly defined in the article, could someone explain what that might be?

How is it so different from other algorithms used for a variety of AI applications

Finally, I can't even guess where they would apply that in a military operation.

If it's things like "Whats best way to attack XXX", or "Should I attack XXX now or tommorrow", I cant see how a tree could deal with all the variables, some totally subjective and not numeric.

Quote

01-22-2019 , 05:32 PM

#10

Federer20

journeyman

Join Date: Jun 2007 Posts: 381

So our beloved game is being used to slaughter people in the future? That is awesome!

Quote

01-22-2019 , 06:13 PM

#11

gamboneee

Carpal \'Tunnel

Join Date: Jun 2010 Posts: 9,163

Quote:

Originally Posted by PTLou

If it's things like "Whats best way to attack XXX", or "Should I attack XXX now or tommorrow", I cant see how a tree could deal with all the variables, some totally subjective and not numeric.

how would a room full of people have a better idea? it can process immense amounts of data; weather, positions, previous encounters, run sims, provide different outcomes, and let someone pick which.

remember all the war room meetings in GoT? imagine the computer watching the show 1 million times, playing each battle each possible way another 100k times, and then asking it what to do.

Quote:

So our beloved game is being used to slaughter people in the future? That is awesome!

just ask snowden (great movie also)

Quote

01-22-2019 , 07:58 PM

#12

PTLou

JellyBean

Join Date: Nov 2008 Posts: 5,516

Quote:

Originally Posted by gamboneee

how would a room full of people have a better idea? it can process immense amounts of data; weather, positions, previous encounters, run sims, provide different outcomes, and let someone pick which.

Ok, that helps.

But large scale data mining algorithms have been around for while.

Simulators as well.

Hardware (and algorithm) advances have certainly made those better, faster over the years.

What is the new function that this Strategy Robot provides. Are you saying its doing the same stuff, just able to handle more input and/or do it faster?

So is Robot Strategy just throughput enhancement, and not any new AI algorithms.

I ask only for two reasons.

1) Im interested and want to understand this at high level.

2) I need to know how much time we have before this Robot thingy becomes self aware and we all die.

Quote

01-22-2019 , 08:13 PM

#13

gamboneee

Carpal \'Tunnel

Join Date: Jun 2010 Posts: 9,163

because it learns. you are describing putting 3x3 into a calculator over and over. grinding out ONE answer to ONE question. instead of compiling all answers, categorizing, adjusting based on answers, rinse repeat, as fast as it can until the best solution is found. it can see patterns, tendencies, etc along with theory games outcomes to decide whats best.

military application would take a huge overhaul and who knows if they get it right or not. but the tech could be easily enough applied to anything that needs a decision

Quote

01-22-2019 , 08:22 PM

#14

PTLou

JellyBean

Join Date: Nov 2008 Posts: 5,516

thanks.

I need to ponder this some more.

But back to one of my original questions.

The article painted a picture that this mysterious Strategy Robot was something new. What is it doing that all the existing technology is not?

Quote

01-22-2019 , 08:45 PM

#15

nuggz

enthusiast

Join Date: Sep 2016 Posts: 77

Quote:

Originally Posted by fishfood69er

"now shoot our own soldiers 46 percent and the enemies soldiers 53 percent of the time, we must remain balanced here. others wise were too exploitable. "
-libratus

instead of allahuakbar, suicide bombers are now going to be shouting "#**** GTO!"

Quote

01-22-2019 , 08:50 PM

#16

gamboneee

Carpal \'Tunnel

Join Date: Jun 2010 Posts: 9,163

Quote:

Originally Posted by PTLou

it learns... like i said above. machine learning is the only reason the idea to start these bots exists. hence being more than a calculator, it doesnt just grind out answers one by one, it uses each outcome to influence the next, over and over and over

Quote

01-22-2019 , 08:57 PM

#17

PTLou

JellyBean

Join Date: Nov 2008 Posts: 5,516

But machine learning is nothing new, actually sort of old.

What's new?

Quote

01-23-2019 , 01:30 AM

#18

JackBurton

grinder

Join Date: Jul 2010 Posts: 588

In one of the videos on youtube about Libratus and its authors, one of the comments made by some tech enthusiast was that this was going to be used to control people in the future, like it was a good thing. And that comment had many thumbs up (maybe irony?)

Or they are just bots scraping social media promoting themselves, who knows...

Quote

01-23-2019 , 11:54 AM

#19

RussellinToronto

veteran

Join Date: Oct 2007 Posts: 2,198

Quote:

Originally Posted by Federer20

So our beloved game is being used to slaughter people in the future? That is awesome!

1950, an early application of von Neuman.

Quote

01-24-2019 , 10:38 PM

#20

PTLou

JellyBean

Join Date: Nov 2008 Posts: 5,516

Deepmind beats Starcraft pros.

https://www.theverge.com/2019/1/24/1...raft-2-victory

Quote

01-25-2019 , 10:04 PM

#21

SolarAU

old hand

Join Date: Apr 2015 Posts: 1,795

Libratus, or rather the AI architecture behind it is capable of much more than solving poker same with other AI projects such as deepmind, the hurdle is correctly implementing and applying this technology to these military applications.

Is it just me or is there a real Skynet vibe going on. Judgement day is coming boys

Quote

01-28-2019 , 03:43 AM

#22

rakemeplz

Carpal \'Tunnel

Join Date: Feb 2006 Posts: 12,648

Quote:

Originally Posted by PTLou

Deepmind beats Starcraft pros.

https://www.theverge.com/2019/1/24/1...raft-2-victory

I mean this is sort of unimpressive, strategicly, if you assume the computer can "micro" much better than a top human player. It appears it could, if I take the artcile you posted at face value.

Quote

01-28-2019 , 04:25 AM

#23

PTLou

JellyBean

Join Date: Nov 2008 Posts: 5,516

Quote:

Originally Posted by rakemeplz

I mean this is sort of unimpressive, strategically, if you assume the computer can "micro" much better than a top human player.

could you explain what you mean in bolded above?

Also, do you have any guess about the "Strategy Robot" mentioned in the original Libratus in OP?

Gamboree was helpful, but the algorithms / heuristics he mentioned have been around for quite a while (of course always getting faster with better hardware/software)

But, the article doesn't leave me the impression that Strategy Robot is just faster.

It comes across as a new algorithm, or is putting existing methods together in a totally new way.

OR

It is pretty much same as existing tech , maybe a bit faster in heuristics or something else, and Strategy Robot was used simply to dramatize the news.