• Welcome to Autism Forums, a friendly forum to discuss Aspergers Syndrome, Autism, High Functioning Autism and related conditions.

    Your voice is missing! You will need to register to get access to the following site features:
    • Reply to discussions and create your own threads.
    • Our modern chat room. No add-ons or extensions required, just login and start chatting!
    • Private Member only forums for more serious discussions that you may wish to not have guests or search engines access to.
    • Your very own blog. Write about anything you like on your own individual blog.

    We hope to see you as a part of our community soon! Please also check us out @ https://www.twitter.com/aspiescentral

Artificial Intelligence. Anyone?

Canismajoris

Hypergiant
AI is something I have been interested for a long time (+20years or so). My interest started from science fiction, but is now maturing to real applications. I am not an expert by any means, but seek to learn more.

So I wondered if ther are others who would be interested in this topic?

Ps. AI ethics / HLMI visions are cool to ponder, but I would rather talk about concrete applications, places to study AI online, news and stuff like. <3
 
I suppose it interests me to some degree, though as gaming is my main hobby, it tends to focus around that. I apologize if this isnt as interesting as other applications of AI, but I may as well talk a bit about it anyway.

However, most games are notorious for having AI that's dumber than a sack of hammers. Many games that feature competitive multiplayer offer the option of playing against "bots" instead, AI opponents that are designed to fill in the spot that normally would be occupied by an actual player. Usually, they are abysmal. Zero ability to adapt, braindead "strategies", and a lack of creative... well.... anything. As a rule, AI opponents in most games are considered very easy to beat.... unless they are given the ability to cheat, which they frequently are. Even amongst professional AAA developers, AI just never manages to be any good.

There is one, and only one, exception that I'm aware of. A particular developer, who I know personally, tends to make alot of strategy games, but specializes in the creation of AI, to the point where it puts basically everyone else to shame. Their flagship game, called AI War, is an RTS (real time strategy) game that has the very rare trait of not being a multiplayer competitive game... it pits the player against an AI, never a human opponent.... for it does not need one. I dont know how they do it, but in all my years of gaming, it's the one AI that is capable of constantly being a true threat. It is creative, it adapts, it will frequently use surprising strategies. Many players like to write up tales of recent playthroughs, where the AI did unexpected and very effective stuff.... even those players that have put 1000 hours into the game still have new tales to tell. What's more, there isnt just ONE AI... there are a great many to choose from, each with it's own traits, many of which also bring some unique units to the war, increasing the complexity that they need to deal with.

But even more ridiculous, the game doesnt play like any other RTS. In an RTS, the idea is usually simple: You play on a single map, each player starts in their respective base, and the idea is to use your army to capture resources and eventually destroy the opposing base. Well, in this game.... there are 80 maps. At the same time (and the whole mess is procedurally generated). With *thousands* of units roaming them. The maps are not "paused" simply because you are not looking at them... they all run all at the same time, and units travel between them constantly (fortunately though, you can pause the game at any time to issue orders... it would be nigh-impossible otherwise). The simulation is *massive*. Going even further, the game is hideously complex... I often call it the Dwarf Fortress of RTS games. If anyone knows what Dwarf Fortress is... I doubt I need to explain further. It took me a LONG time to learn how to play AI War. I normally never, ever struggle with learning a new game. But in this case, I watched a good 30 tutorial vids to learn it. Even after learning the game and practicing, the difficulty remains very brutal. Most players will get utterly stomped many, many, MANY times before even coming close to a win. This simply doesnt happen in other games in the genre... RTS games are normally known for even-more-braindead-than-usual AI that is extremely easy to exploit and defeat, even for new players.

All of this means an AI that is utterly beyond any other game I've seen in all of my years as a gamer. The thing fascinates me to no end. It fascinates the entire fanbase, too.


Yeah, I know, many would say "bah, it's just a game", but still. I find it amazing nonetheless. As far as I'm concerned, it's one hell of an achievement... particularly considering that this is a SMALL developer. The horrid monstrosity was developed by TWO people. Not a team of hundreds.
 
So I wondered if ther are others who would be interested in this topic?
This is why I'm studying computational linguistics. I want to work on spoken dialog systems intended to catch subtext a human might not, or notice when someone's in crisis. We just had a speaker this week that's been working with a crisis text service to figure out how the most successful crisis counselors are helping people and to develop a training simulation from that.
 
I have a growing curiosity around this myself. My earliest exposure to it was through video games and through songs and literature that depict dystopian AI-ridden futures. But now that its becoming an ever increasing reality, the moral side of things is where I'm really... contemplative. There doesn't seem to be any straightforward answer to the moral questions of where it is headed. And its incredibly multi-layered. Like the OP I'm no expert, but I'm eager to understand and learn more.
 
Since you're interested in the real-life applications of AI, I think you would probably find interesting topics on some company blogs. I was just reading an article about AI and food design a few days ago on the Technology Innovation blog from Accenture, and I'm sure Google, IBM and a host of tech start-ups all maintain blogs to showcase the possible applications AI offers.

From there, you'll be introduced to more topics and can maybe pick a few to dive deeper into: maybe machine-learning, Internet of Things, or blockchain to begin with, and then I have no doubt that you could fall in a rabbit hole of information ;)
You could also refine your searches by looking up a particular field of application: maybe another of your interest is medicine, or languages, or food, or finance... every field has possible applications, so you might as well pair interests at first, don't you think?
 
most games are notorious for having AI that's dumber than a sack of hammers.
I think that there is no actual AI running in those stupid games: atleast if we agree to think that AI should be learning algorithm?

I would also guess, that your friends game uses somekind of learning system that helps the game to predict the moves better. If also the learnt knowledge is stored, the game should get even harder and harder when the AI learns more.

And yes, games are one of the many practical environments to implement AI. There are rules that make the limits and therefore there is a system to make predictions: that is what AI is truly about at the moment. :)
 
This is why I'm studying computational linguistics. I want to work on spoken dialog systems intended to catch subtext a human might not, or notice when someone's in crisis. We just had a speaker this week that's been working with a crisis text service to figure out how the most successful crisis counselors are helping people and to develop a training simulation from that.
Very interesting! (I actually study communication management besides of my work and we just had a course about crisis communication.)

But for the application: I assume you would be using some kind of facial recognizion system that tries to find patterns from big data that indicate that there is happaning more than eye can catch? I would assume that the hard part is to be able to link emotions and facial muscle movements. This would require a studio, where faces are shot on video while emotions are tried to be invoked. This must not be nothing new, but to make this system good enough, it might require more resources than are available? Or how you plan to teach your AI?
 
I think that there is no actual AI running in those stupid games: atleast if we agree to think that AI should be learning algorithm?

I would also guess, that your friends game uses somekind of learning system that helps the game to predict the moves better. If also the learnt knowledge is stored, the game should get even harder and harder when the AI learns more.

And yes, games are one of the many practical environments to implement AI. There are rules that make the limits and therefore there is a system to make predictions: that is what AI is truly about at the moment. :)

He's actually written up explanations on how parts of it work.... frankly I understood very little of it. If I recall correctly, he's done a few presentations too.

The one bit I did grasp though: The main AI has many "sub" AIs that run at the same time. The main one acts as a commander/leader, handling all of the meta-strategy (aka, the truly complex part that spans the galaxy map). Orders are relayed to lesser "squadron leaders" (which dont have an actual ship presence, they're more a logic blob that follows a group of ships around). And then those give smaller orders to each individual ship in their group, which have miniature AIs to carry out individual orders (this level of AI is also active in player-controlled ships).

Considering the sheer ridiculous number of units that a typical game of AI War involves, I dont understand how any of this functions without the computer violently detonating.

The one other thing I do know is that the different AIs have certain restrictions/rules on their behavior. There's one AI called "the tank" that will always focus on high defense and "turtling" strategies, for instance. Another one (I cant remember the name) will choose a specific ship type (and there's like over 100 of those) and have a tendency to put a very high focus on using that ship. There are ALOT of these AI "personalities" to choose from, and they frequently added more. Doesnt matter which ones you pick for a given playthrough: They will still manage to do surprising things and beat your face in.

The other odd part: You actually choose TWO AIs, not one, when starting a new game. So that's two core "commanders" each with their own massive fleet and set of controlled planets. They do not fight each other and instead work together against you. Despite them not sharing the same "personality" type (well, unless you select two of the same one).
 
I have a growing curiosity around this myself. My earliest exposure to it was through video games and through songs and literature that depict dystopian AI-ridden futures. But now that its becoming an ever increasing reality, the moral side of things is where I'm really... contemplative. There doesn't seem to be any straightforward answer to the moral questions of where it is headed. And its incredibly multi-layered. Like the OP I'm no expert, but I'm eager to understand and learn more.
The best book I have read about AI in general and AI ethics is "Superintelligence". It goes REALLY deep and to be honest, is not the easiest book to understand. BUT I do recommend it if you are willing to dive in the deep end of the pool.

https://www.amazon.com/Superintelligence-Dangers-Strategies-Nick-Bostrom/dp/1501227742
 
Since you're interested in the real-life applications of AI, I think you would probably find interesting topics on some company blogs. I was just reading an article about AI and food design a few days ago on the Technology Innovation blog from Accenture, and I'm sure Google, IBM and a host of tech start-ups all maintain blogs to showcase the possible applications AI offers.

From there, you'll be introduced to more topics and can maybe pick a few to dive deeper into: maybe machine-learning, Internet of Things, or blockchain to begin with, and then I have no doubt that you could fall in a rabbit hole of information ;)
You could also refine your searches by looking up a particular field of application: maybe another of your interest is medicine, or languages, or food, or finance... every field has possible applications, so you might as well pair interests at first, don't you think?
Good advice in general. I have taken one free university course about basic AI (it included the basic concepts of machine learning.) then I also have read some blogs and also some McKinseys reports, they are pretty cool. All in all the next level I plan to take is to enroll into some online course maybe in Coursera or Usemy. In Coursera the courses held by Andrew Ng are told to be super good!
 
He's actually written up explanations on how parts of it work.... frankly I understood very little of it. If I recall correctly, he's done a few presentations too.

The one bit I did grasp though: The main AI has many "sub" AIs that run at the same time. The main one acts as a commander/leader, handling all of the meta-strategy (aka, the truly complex part that spans the galaxy map). Orders are relayed to lesser "squadron leaders" (which dont have an actual ship presence, they're more a logic blob that follows a group of ships around). And then those give smaller orders to each individual ship in their group, which have miniature AIs to carry out individual orders (this level of AI is also active in player-controlled ships).

Considering the sheer ridiculous number of units that a typical game of AI War involves, I dont understand how any of this functions without the computer violently detonating.

The one other thing I do know is that the different AIs have certain restrictions/rules on their behavior. There's one AI called "the tank" that will always focus on high defense and "turtling" strategies, for instance. Another one (I cant remember the name) will choose a specific ship type (and there's like over 100 of those) and have a tendency to put a very high focus on using that ship. There are ALOT of these AI "personalities" to choose from, and they frequently added more. Doesnt matter which ones you pick for a given playthrough: They will still manage to do surprising things and beat your face in.

The other odd part: You actually choose TWO AIs, not one, when starting a new game. So that's two core "commanders" each with their own massive fleet and set of controlled planets. They do not fight each other and instead work together against you. Despite them not sharing the same "personality" type (well, unless you select two of the same one).
Wow! This game sounds pretty awesome! If only I had time to play...the last time I played was on my winter holiday when we completed the Secret of Monkey Island with my wife who loves that game.. Talking about AI! xD
 
What limited intelligence I have is real - not crazy about the artificial kind. tee hee
And people can always tell...!
full
(It isn't real; it's store-bought...)
 

New Threads

Top Bottom