Support NeoGAF

[Ozzy Onya A2Z]

ITT gamers don't realise they already use APIs from Google, Azure, AWS with neural networks, general and specific AI functions as well as custom developments. They're all heading that direction and gaming will have some innovations and borrow from others just as the rest will.

 

[Warablo]

Then how come AI seems like its getting worse as the years go on?

EDIT: I am talking about video game AI

 

[Three]

Gaming is a crazy way of getting behavioural data that much I can agree with. You can create any situation and monitor peoples response or behaviour.

 

[ItsTheNew]

Then how come AI seems like its getting worse as the years go on?

Ai is getting REALLY good at gathering info about you and selling it to every company under the sun.

 

[Shadowplay1979]

you just need to ask for an enemy capable of beating Data, I mean, Sherlock.

I cant wait for the moriarty ai that becomes the best mario cart opponent

 

[Kev Kev]

 

[Sentenza]

No it's not.

All this shows is your severely limited understanding of what we call "AI" in videogames.

"AI" in videogames isn't actually AI at all. It does next to no learning of anything and in the vast majority of cases comprises simple pathfinding and simulated behavioural algorithms that haven't meaningfully changed in the past 30 years.

ML, DL, Neural Networks are the domain in which any meaningful computer intelligence might emerge, and these fields barely find any utilisation in videogames beyond limited examples like graphical tricks to improve image quality like DLSS, and offline procedural generation of art assets to help speed-up the art production workflow of the dev process.

The subroutines that determine enemy and NPC behaviour in games are simple non-learning functions that take in a limited number of inputs and spit out the exact same pre-canned behavioural outputs regardless of the contextual backdrop of the game state. It's not AI.

What games do today is completely irrelevant, though?
Seems like you are missing the point.
There's a reason if both Open AI and Alpha Zero have been trained with playing board games and videogames of any sort, and that's that they offer an environment self-contained enough to have a large but finite number of variables while still offering enough complexity to require flexible approach without a chance to brute-force through them.

On a side note, anyone with even just a passing interest on the topic should check Two Minutes Papers on youtube: https://www.youtube.com/user/keeroyz
Monitoring upclose the uncanny rate at which this field is improving feels almost unreal.
We are basically reaching monthly goals that were thought to be decades away barely few years ago.

 

[ThatGamingDude]

So what you're saying is SHODAN is probably really into QUAKE Champions??

 

[TheThreadsThatBindUs]

ITT gamers don't realise they already use APIs from Google, Azure, AWS with neural networks, general and specific AI functions as well as custom developments. They're all heading that direction and gaming will have some innovations and borrow from others just as the rest will.

While true, this has no bearing on the OP's hypothesis that gaming as a medium is inherently better suited to facilitating the emergence of Artificial General Intelligence.

The primary argument that most posters here have been raising is that gaming is not the best medium for this outcome to eventuate. If/when artificial general intelligence occurs, it will most likely emerge in more specifically scientific fields of study dedicated to AI development. Not by accident, by a NN that watched a lot of gamers playing Call of Duty.

What games do today is completely irrelevant, though?
Seems like you are missing the point.
There's a reason if both Open AI and Alpha Zero have been trained with playing board games and videogames of any sort, and that's that they offer an environment self-contained enough to have a large but finite number of variables while still offering enough complexity to require flexible approach without a chance to brute-force through them.

On a side note, anyone with even just a passing interest on the topic should check Two Minutes Papers on youtube: https://www.youtube.com/user/keeroyz
Monitoring upclose the uncanny rate at which this field is improving feels almost unreal.
We are basically reaching monthly goals that were thought to be decades away barely few years ago.

Why is it irrelevant?

If anyone is missing the point it's you, as your post is tangential to the core discussion of this thread. You're not even making a point in support of the OP's argument. Your post simply amounts to a statement of the following, "AI development is happening quickly and the current state of the art has used games with simple defined rulesets to train NNs".

Well yeah, sure! NNs have been trained using simple games like Chess, Mario and even more complex FPS games with player movement and interaction in 3-dimensions.

Does that mean those same NN are somehow going to evolve AGT? To try to claim an affirmative response to this question would be a leap of logic that defies all reason or evidentiary justification. And yet that is essentially what the OP is trying to argue.

The point being argued by the OP isn't about how fast the field of AI is developing, nor even whether games have utility in that development process. The's OP's argument is that gaming is somehow inherently the best possible and most likely medium in which Artificial General Intelligence will emerge.

There is simply no evidence available to support this view. And posters like myself have presented the primary counter-argument that gaming isn't even currently the field in which the most state of the art AI development advancement is currently happening, so what reason is there to believe it will be so in the future?

 

[Sentenza]

Why is it irrelevant?

because the point was never "Top level advanced AI is already IN games today.
It's that games and their complex systems of rules and variables are often being used to train and put to test AI.

The main difference of Alpha Zero from its predecessor Alpha GO was precisely its capability to "learn" and adapt to countless different games and rule systems it wasn't specifically coded for.

You're not even making a point in support of the OP's argument.

But I wasn't trying to support the Op's argument. In fact I think the OP's argument is stupid as hell.
I was specifically dismissing yours, instead.

 

[TheThreadsThatBindUs]

because the point was never "Top level advanced AI is already IN games today.
It's that games and their complex systems of rules and variables are often being used to train and put to test AI.

This clearly isn't the point being made by the OP, and therefore this was not the point my reply was responding to.

The main difference of Alpha Zero from its predecessor Alpha GO was precisely its capability to "learn" and adapt to countless different games and rule systems it wasn't specifically coded for.

Again, tangential to the discussion I was having with the OP.

But I wasn't trying to support the Op's argument. In fact I think the OP's argument is stupid as hell.
I was specifically dismissing yours, instead.

Well you failed, because the point you're dismissing isn't one I've even made. You're misconstruing my argument and trying to dismiss it. That's called arguing a strawman.

The original post of mine you were responding to was a response to the OP's hypothesis. All you've done is move the goal posts by changing the argument entirely to dismiss my counter-argument to the OP.

All you've managed to do is waste both of our time.

 

[Sentenza]

Well you failed, because the point you're dismissing isn't one I've even made. You're misconstruing my argument and trying to dismiss it. That's called arguing a strawman.

You were talking about the low complexity of AI in modern games and how little it progressed in the last few years.

I'm telling you that "the AI in current games" is not particularly relevant. Advancement in modern AI (i.e. recognizing, learning, exploiting and perfecting patterns with additional iterations) aren't being made by putting it into games, to begin with, but by making it play these games instead.
See those public exhibitions with Open AI playing DOTA 2 against a single player at first and then against full teams few months later.

 

[TheThreadsThatBindUs]

You were talking about the low complexity of AI in modern games and how little it progressed in the last few years.

I rest my case. You missed the point completely.

This was not the subject of the discussion nor the point I nor the OP was making.

 

[Victor Omega]

I love the progression of this thread. It starts as an insane idea then thought-provoking it's could be implemented then this:

Sony patents enemy AI learning that adjusts in real time for difficulty​

https://www.neogaf.com/threads/sony...-adjusts-in-real-time-for-difficulty.1589465/

 

[mxbison]

Isn't all the this "AI" we have still only machines learning by repeating some process and adjusting the expected result values?

Doesn't sound like any form of intelligence to me.

 

[Woo-Fu]

Anybody who has played games for years doesn't believe this. If anything the AI in games is going downhill, on average. Your dog is more likely to accidentally find a cure for cancer.