Why hasn't AI improved in the same way that e.g., graphics has?

[Wizzington]

Sooooo Wiz while you are here and we have your attention, are you really working on the AI?

Unless I've been imagining the last few months, I sure am.

 

[Alerias]

Unless I've been imagining the last few months, I sure am.

Keep it up! And no pressure, but your work is the biggest variable in just how much a step forward the game will be compared to 5.2

 

[Sn3ipen]

So Wiz, whats your take on this topic? Its always interesting to hear devs talk about such things. To hell with it, you and Shams could do an whole episode in his show on Twitch about the subject.

...and btw, congratulations with the job.

 

[FOARP]

So Wiz, whats your take on this topic? Its always interesting to hear devs talk about such things. To hell with it, you and Shams could do an whole episode in his show on Twitch about the subject.

...and btw, congratulations with the job.

Seconded - why hasn't AI improved at the the same rate processing power has? Or is this just my impression?

 

[Wizzington]

So Wiz, whats your take on this topic? Its always interesting to hear devs talk about such things. To hell with it, you and Shams could do an whole episode in his show on Twitch about the subject.

...and btw, congratulations with the job.

I'm going to be brief since I'm working but if I had to name one reason that AI hasn't seen the kind of steady improvement we've seen in other game mechanics I would chalk it down to lack of planning. In my opinion, a good AI needs to designed along with the game and accounted for on all stages. Too often, AI is something that is thrown on after the features are already long finalized (rarely with sufficent time budgeted to do a reasonably good job at it) and so you end up with a feature that is fully realized to the player but only understood by the AI on a shallow surface level.

Processing power should not be discounted, but I don't think it's the primary problem. With enough time and skill, you could make strategy AIs vastly superior to the ones we see today, but without proper planning you end up wasting a lot of that time and skill. The way I see my job here at Paradox, besides doing the actual AI programming, is to be the 'voice of the AI' during the design process, commenting and making suggestions from its perspective. Of course I'm not in charge of feature design or anything like that but no matter how good your game designers are (and ours are very very good), it still always helps to have someone whose sole responsibility is to back the AI's corner, so to speak.

 

[Chromos]

Nicely written.

 

[FOARP]

I'm going to be brief since I'm working but if I had to name one reason that AI hasn't seen the kind of steady improvement we've seen in other game mechanics I would chalk it down to lack of planning. In my opinion, a good AI needs to designed along with the game and accounted for on all stages. Too often, AI is something that is thrown on after the features are already long finalized (rarely with sufficent time budgeted to do a reasonably good job at it) and so you end up with a feature that is fully realized to the player but only understood by the AI on a shallow surface level.

Processing power should not be discounted, but I don't think it's the primary problem. With enough time and skill, you could make strategy AIs vastly superior to the ones we see today, but without proper planning you end up wasting a lot of that time and skill. The way I see my job here at Paradox, besides doing the actual AI programming, is to be the 'voice of the AI' during the design process, commenting and making suggestions from its perspective. Of course I'm not in charge of feature design or anything like that but no matter how good your game designers are (and ours are very very good), it still always helps to have someone whose sole responsibility is to back the AI's corner, so to speak.

This sounds, in a geeky software-dev-talk way (I'm a non-software person working in an IT firm so I often listen to talk about 'Scrum methodology' and so-forth that goes straight over my head), awesome. I look forward to seeing how this approach turns out in future games.

 

[Sn3ipen]

I'm going to be brief since I'm working but if I had to name one reason that AI hasn't seen the kind of steady improvement we've seen in other game mechanics I would chalk it down to lack of planning. In my opinion, a good AI needs to designed along with the game and accounted for on all stages. Too often, AI is something that is thrown on after the features are already long finalized (rarely with sufficent time budgeted to do a reasonably good job at it) and so you end up with a feature that is fully realized to the player but only understood by the AI on a shallow surface level.

Processing power should not be discounted, but I don't think it's the primary problem. With enough time and skill, you could make strategy AIs vastly superior to the ones we see today, but without proper planning you end up wasting a lot of that time and skill. The way I see my job here at Paradox, besides doing the actual AI programming, is to be the 'voice of the AI' during the design process, commenting and making suggestions from its perspective. Of course I'm not in charge of feature design or anything like that but no matter how good your game designers are (and ours are very very good), it still always helps to have someone whose sole responsibility is to back the AI's corner, so to speak.

Thanks for answering. This sounds like a great way of solving the problem and it is good to see that PDS are taking AI seriously.