Sell me an artefact ... you dont have enough gold ... how much you i have ? .. i dont know .... really nice ai :D :.. :D:D :.. you are poor ... yeah ? how much money do I have ? ... i dont know .... some logarithmical code just told me to look into your status and copy defined value as an argument for answer :D ..
Awesome! :) What is this based on? (You are somehow extracting info from sentence ... and each agent is making his knowledge from it ... and is able to put it into a sentence)
What language are You programming it in (i presume some declarative like Prolog)?
They should put more words into the game that mean the same thing that the game can recognize as having the same values. it isn't needed but it would add more flavor and make it seem more life-like. i know I'd find more amusement being able to walk up to the guy and say how much for a magical token and see him tell me it's 5 gold. just like i would find amusement walking up and saying 'oh shit whats up son?' and seeing him say 'chillin boi, chillin.'
That is really interesting. I love the way that the diffrent options are presented. It makes that kind of text adventure interface viable without punishing the user for not magically knowing how the programmer intends you to phrase things.
@TheReasonWhyGuy I'm not working on this exact game, but still working on logic-based game AI.
Yes, grammar is really hard. It's all interrelated. You change one place and another place breaks. The whole "interactive parser" interface is a way to make up for the fact that the grammar is very limited.
@MartinMagnusson Probably a stupid question but could you evolve grammar?
That would most likely be a slow progress of creating a dictonary and checking sentences for grammar-wise correctness.
In a further step, the actual content could be rated aswell...
Doesn't need to be perfect, but maybe, by some randomness-luck, you get some better results than you have in that video (which is very impressive, already)
@Kram1032 It seems evident that one /could/, given our own evolutionary history. Do you have any ideas for how to do it though? It would make a fantastic YouTube video, watching a time lapse of the creatures learning how to speak!
@MartinMagnusson Heh, well, it even takes humans a good portion of their lives to learn a language. And that in a critical stage, where they are in fact specialized in this kind of learning.
If you don't learn at least one language to a certain extend very early (in the first three years of your life), you wont be able to properly learn that or any other language later.
And you never quite stop learning language. Occationally, you'll hear new words...
What I'd try to do (but I don't know if that's any useful), is to have multiple layers in a dictonary. That is "letter", "word", "phrase"
Every element, aswell as the order of multiple elements would have a frequency-messure combined with them.
Also, a neural net will try to find an abstract meaning behind them. It should learn by human feedback of "bad grammar", "bad logic", "out of context" or a response, which could also be /conversation_end or something.
Additionally, it would be interesting if NPCs could understand each other and even figure out, who is meant in a multi-agent-chat.
e.g. build a couple of independend chat-bots that try to communicate with each other (a supervisor finds out wether the conversation is anywhere near useful), aswell as a couple of human agents.
A conversational start could be given explicitly, e.g. @ChatBotX, but if you talk to multiple agents with different topics...
...it should figure out if it's meant or somebody else.
interrupting a dialogue or joining it should be trained too. The supervisor(s) should be able to tell an inactive chatbot, to figure out ongoing topics and then join (one of) the conversation(s) in a meaningful way.
Phew, that's it.
A project like that should be very complex and slow-paced. Timelapse? phew, lots of hours to compress <.< :)
"How to" says:..."only sentences licensed by the game's grammar are accepted by your character"..."parser will help guide you towards licensed sentences by presenting a list of world alternatives"...
Read it by yourself,i cant copy whole "how to" here.
THE GAME IS VERY SIMPLE.And it seems that you couldnt beat it .
A very interesting experiment, although i am unsure how popular it would be in the gaming market, it makes for a unconventional angle on puzzle challenges. Personally I would try to avoid tasking the player with repetitive inquiries. Hope you succeed with this.
if you asked a guy if he knew something, and then he said no, and a second later you told him that thing, would he know what that thing was for next time you asked him? (or can they learn)
Not in this prototype. But my second video "Intelligent and proactive game characters" shows off a newer version where the characters can learn from the player. In the next version I hope to create characters that learn from players but also stop trusting players that lie to them!
AI will adapt to its surroundings like my new programs will act differnetly if there is ceran files around it. True AI can learn but basic AI can adapt
Artificial intelligence is where the computer can respond to input based on a logical choice. Artificial life is where creatures can learn and adapt to an environment. This seems like it is aimed for a combination of the two whereas the computer can learn and make logical choices with input.
Can not find this anywhere.
glenfoxh 2 days ago
@glenfoxh You'll find it at my web page (which is my name dot com)
MartinMagnusson 2 days ago
PROTIP: GOogling andi land does NOT come up with this.
23ABoyd23 3 weeks ago
@23ABoyd23 *MeGusta/Pokerface*
xSakijix 6 days ago
Sell me an artefact ... you dont have enough gold ... how much you i have ? .. i dont know .... really nice ai :D :.. :D:D :.. you are poor ... yeah ? how much money do I have ? ... i dont know .... some logarithmical code just told me to look into your status and copy defined value as an argument for answer :D ..
7cevz9 4 months ago
Went to andiland.c o m thinking I could find more about this. It wasn't what I was looking for xD
backtobasesix 6 months ago
I would pick this in front of CoD at any day.
S1lentStr1ke 7 months ago
This has been flagged as spam show
Awesome! :) What is this based on? (You are somehow extracting info from sentence ... and each agent is making his knowledge from it ... and is able to put it into a sentence)
What language are You programming it in (i presume some declarative like Prolog)?
daEmoNicky 8 months ago
Comment removed
daEmoNicky 8 months ago
i searched the game name in google and saw porn. O.o
mrSanxable 1 year ago
A magical land where everyone talks like yoda... amazing this is
JAFOOOLY 1 year ago 2
@JAFOOOLY the comments owner is you and your thumbs up is +1
mrSanxable 1 year ago 3
They should put more words into the game that mean the same thing that the game can recognize as having the same values. it isn't needed but it would add more flavor and make it seem more life-like. i know I'd find more amusement being able to walk up to the guy and say how much for a magical token and see him tell me it's 5 gold. just like i would find amusement walking up and saying 'oh shit whats up son?' and seeing him say 'chillin boi, chillin.'
erikstevenpierro 1 year ago
This was a painful video to watch. I wasted 5min of my life.
rfengr 1 year ago
really amazin dude !!!
ollyuzr 1 year ago
Honestly, I think I want to try my hands at something like this :)
TheReasonWhyGuy 1 year ago
at the time when he buys the axe I think "use axe to get other items from the guy"
SalsaTiger83 1 year ago
This is awesome it's like a very low scripting language that kids could play
CTSSTC 1 year ago
Link?
88Meters 1 year ago
i did this pretty much just like this vid, and at the end (after about 5 mins) I was like cluster fuck. . .what just happened
rotfmol 1 year ago
The magical token's owner is me and the axe's owner is me and the bread's owner is me.
Wow, awesome conversations! Really interesting stuff!!
ArmoredSandwich 1 year ago
this is impressive.
anzwertree 1 year ago
Use AXE on MAN
rankaratar 1 year ago
what would happen if you say "i shall kill you"?
talkron300 1 year ago
I sold the axe to Tulku then bought it back from him earning a tidy 2 gold profit.
ikkonoishi 1 year ago
That is really interesting. I love the way that the diffrent options are presented. It makes that kind of text adventure interface viable without punishing the user for not magically knowing how the programmer intends you to phrase things.
ThatOtherDev 2 years ago
God damn, this is fascinating.
The way Shinu phrased, "the magical token's owner is me, the axe's owner is me, the bread's owner is me" was really amusing.
"Shinu" means "die" in Japanese. Intentional?
LogicalPhallusy 2 years ago
@LogicalPhallusy Really? Not intentional, but maybe one could build a background story for the game based on that :)
MartinMagnusson 2 years ago
@MartinMagnusson Are you still working on this game?
Did you find English grammar difficult to program?
It seems in english, there are 1000 ways to say 1 thing :|
TheReasonWhyGuy 1 year ago
@TheReasonWhyGuy I'm not working on this exact game, but still working on logic-based game AI.
Yes, grammar is really hard. It's all interrelated. You change one place and another place breaks. The whole "interactive parser" interface is a way to make up for the fact that the grammar is very limited.
MartinMagnusson 1 year ago
@MartinMagnusson Hmm :)
Yeah I'm a programmer, who is also working on an artificial language.
Though honestly, if you want a language good for speaking to computers using, try lojban.
TheReasonWhyGuy 1 year ago
@MartinMagnusson Probably a stupid question but could you evolve grammar?
That would most likely be a slow progress of creating a dictonary and checking sentences for grammar-wise correctness.
In a further step, the actual content could be rated aswell...
Doesn't need to be perfect, but maybe, by some randomness-luck, you get some better results than you have in that video (which is very impressive, already)
Kram1032 5 months ago
@Kram1032 It seems evident that one /could/, given our own evolutionary history. Do you have any ideas for how to do it though? It would make a fantastic YouTube video, watching a time lapse of the creatures learning how to speak!
MartinMagnusson 5 months ago
@MartinMagnusson It certainly would be a lot of work...
There already exist some projects that might at least be useful reference as a starting point, but none of them is quite exactly this...
One example would be 20Q (20Q dot net), which tries to guess what you think. That could be a basis for "understanding"
An other example, Cleverbot (dot com), simply produces likely answers to input questions.
Something that combines both and your ANDIs could lead to something usable...
Kram1032 5 months ago
@MartinMagnusson CONTINUED
I don't have experience in doing anything like that, myself, though.
I just searched the web for interesting stuff in that direction.
Eventually, I want to try to do some neuronal net- or genetic program, myself...
Apparently, the hardest part (in case of GA) is to decide on a genetic code and a proper fitness function...
Evolving understanding AND responding in a human language would only work by supervised learning, I guess...
Kram1032 5 months ago
@Kram1032 Sounds like a good life-long project ;)
MartinMagnusson 5 months ago
@MartinMagnusson Heh, well, it even takes humans a good portion of their lives to learn a language. And that in a critical stage, where they are in fact specialized in this kind of learning.
If you don't learn at least one language to a certain extend very early (in the first three years of your life), you wont be able to properly learn that or any other language later.
And you never quite stop learning language. Occationally, you'll hear new words...
So it's probably a couple of lives long <.<
Kram1032 5 months ago
@MartinMagnusson Continued2
What I'd try to do (but I don't know if that's any useful), is to have multiple layers in a dictonary. That is "letter", "word", "phrase"
Every element, aswell as the order of multiple elements would have a frequency-messure combined with them.
Also, a neural net will try to find an abstract meaning behind them. It should learn by human feedback of "bad grammar", "bad logic", "out of context" or a response, which could also be /conversation_end or something.
Kram1032 5 months ago
@MartinMagnusson Continued3
Additionally, it would be interesting if NPCs could understand each other and even figure out, who is meant in a multi-agent-chat.
e.g. build a couple of independend chat-bots that try to communicate with each other (a supervisor finds out wether the conversation is anywhere near useful), aswell as a couple of human agents.
A conversational start could be given explicitly, e.g. @ChatBotX, but if you talk to multiple agents with different topics...
Kram1032 5 months ago
@MartinMagnusson continued4
...it should figure out if it's meant or somebody else.
interrupting a dialogue or joining it should be trained too. The supervisor(s) should be able to tell an inactive chatbot, to figure out ongoing topics and then join (one of) the conversation(s) in a meaningful way.
Phew, that's it.
A project like that should be very complex and slow-paced. Timelapse? phew, lots of hours to compress <.< :)
Kram1032 5 months ago
woah, really cool! good job!
ADeadRock 2 years ago
Thats some interesting idea for a game
*thumbs up*
if you ever search for a pixel artist, then tell me please
maybe we could work together =)
to see its work in such great projects is a great feeling i bet ^^
PunikaTV 2 years ago
Question: Tulku what is your wealth?
Awnser: My wealth is 6gold.
My Thought: i have 0 gold, an axe, no cops around me, a wealthier feeble robot with no means of self-proctection..
My Action: Cyber-MUGGING
JetsuSeal 2 years ago 24
Lol =)
Actually, in the next variant of the game, the player will act as a cop defending against criminals
MartinMagnusson 2 years ago
LoL Cop Vs Civillians another day in our "Normal" world. Now it's soon gonna be the same way in a games lol
JetsuSeal 2 years ago
it seems to work like prolog
blytqb 2 years ago
I'd be surprised if the source isn't at least partially in prolog.
CrazyHorseInvincible 2 years ago
Ok, this isn't really artificial intellegence. It's more logics. (Though they are related)
Chrismonchan 2 years ago 4
These type of games are awesome and open up a world of possibilities.
dandymcgee 2 years ago
this is saurabh shrivastava video
saurabh0708 2 years ago
I tried this, but the AI didn't understand a single word I said.
liquos 3 years ago
RTFM
"How to" says:..."only sentences licensed by the game's grammar are accepted by your character"..."parser will help guide you towards licensed sentences by presenting a list of world alternatives"...
Read it by yourself,i cant copy whole "how to" here.
THE GAME IS VERY SIMPLE.And it seems that you couldnt beat it .
IMMMedved 3 years ago
I want to play that really bad!! That looks very chalanging
hydrolisk1792 3 years ago
cool
sxeeplta 3 years ago
this is great ,
i will try it myself :D
zhgaaaan 3 years ago
This comment has received too many negative votes show
i call bullshit on this
JUKIO01 3 years ago
I tried really hard to do something like this with Perl... epic fail.
Awesome Job!
cabmill 3 years ago 2
Did you make this game? If so GREAT job, I wish I could make a game this cool. What's it built in?
MasterJake777 3 years ago 4
It's built entirely in Lisp. Though I'm working on a new version that will have the AI code in Lisp and the graphics code in C++.
MartinMagnusson 3 years ago
well your very good at it, i really loved it
MasterJake777 3 years ago
The Magical Token's owner is me.
A winner is me.
Lol.
DFaz3 3 years ago 19
Comment removed
IMMMedved 3 years ago
Comment removed
IMMMedved 3 years ago
A very interesting experiment, although i am unsure how popular it would be in the gaming market, it makes for a unconventional angle on puzzle challenges. Personally I would try to avoid tasking the player with repetitive inquiries. Hope you succeed with this.
FrankDodd 3 years ago 3
Where can i get this?
thani94 3 years ago
There's an URL at the end (4:50) of the movie.
MartinMagnusson 3 years ago
Magni: The stupidness of my game is fucking.
smellycorpse 3 years ago
Quality coding and quality choice of music. Can I ask what this track is? Thanks :)
jeromewilson 3 years ago
Thanks!
The track is "Stimulation Two" by Adam Skorupa, which you can buy at shockwave-sound.
MartinMagnusson 3 years ago
Thanks. Keep up the good work.
jeromewilson 3 years ago
if you asked a guy if he knew something, and then he said no, and a second later you told him that thing, would he know what that thing was for next time you asked him? (or can they learn)
touchestoomuch 3 years ago
Not in this prototype. But my second video "Intelligent and proactive game characters" shows off a newer version where the characters can learn from the player. In the next version I hope to create characters that learn from players but also stop trusting players that lie to them!
MartinMagnusson 3 years ago
magni's an idiot
ShowNoMercy77 3 years ago
this isnt ai
its merely if/then statements, as a million others have already said!!
theabyscalls 3 years ago
Uh huh. You've seen the source code, have you, Linus?
jeromewilson 3 years ago
No, true AI can kill astronauts.
DocNinja 4 years ago
No true AI, can change its stategy, understand intimidation and stealth
astromonger 4 years ago
Sorry, but that is not Intelligence.
Those are just mere if-then statements.
True AI can learn.
999madcows 4 years ago 2
Well isnt IF-THEN-ELSE basic ai
AI will adapt to its surroundings like my new programs will act differnetly if there is ceran files around it. True AI can learn but basic AI can adapt
simkinggold101 3 years ago
Artificial intelligence is where the computer can respond to input based on a logical choice. Artificial life is where creatures can learn and adapt to an environment. This seems like it is aimed for a combination of the two whereas the computer can learn and make logical choices with input.
K25125 3 years ago
great job
damianpaz 4 years ago
thumbs up
Fromzon 4 years ago
Looks fantastic!
benjamin314 4 years ago