Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Laida]

Hello.
It is somewhat confusing that robots, which according to the plot, should be very smart, in fact, have AI so dumb as any other robots. I mean, analysts, researchers, tinkerers, key plot characters...

Example 1:

Spoiler (click to show/hide)

Let consider situation when analyst staying near the alien artifact at the Quarantine and see that player breaking fences which surrounding the artifact. Now analyst will only send distress signal and make a stasis beam to intercept player's fleeting. But this is sane only if analyst's allies or Garrison are nearby. In other case, it is more suitable for analyst to try to grab the artifact and flee with it or just to run to the point where he will able to call for reinforcements more effectively. Also I was very surprised when I saw that scavengers utilizes artifacts as every other things... May be, they do this because they are dumb, but more smart robots can interfere in this kind of situations and save artifacts from recycling. 

Example 2:

Spoiler (click to show/hide)

When Warlord assaulting Complex, he go ahead recklessly and can die easily, which is stupid for the leader... I think it will more suitable if he will behave more carefully, especially when at low health.

Example 3:

Spoiler (click to show/hide)

When player discovers the derelict researcher behind the wall in the Caves and destroys the wall, researcher don't stay on place anymore and don't go away to stairs, but starts to roam nearby and become easy target for unaware patrols. It is stupid and sad. 

I saw some conversations about AI at forum but didn't find exactly this kind of suggestion, so opened a new thread about it.

[Kyzrati]

The thing about AI development is that it's an endless rabbit hole. I know because I've spent literally two entire years of my life developing smart AI for a game before. In the end it's not worth the vast amount of work involved and I decided I wouldn't expend lots of resources on diminishing returns like that. The time is far better spent adding other content and features.

I mean yeah if someone wanted to pay me to develop Cogmind for many more years I could do lots of cool stuff, but I have a super long list of great features to add so everything has to be prioritized accordingly. A lot of them will just never happen.

That said, honestly the examples you bring up are minor points when you think about these characters from their point of view, or within the world's lore:

Spoiler (click to show/hide)

• Analysts are still just robots, and as Unaware they have no personalities nor fear for their lives so holding you there while they scan you is just doing their job. You have seen intercepts, right?
• Warlord is not supposed to be smart, he's supposed to be a macho death robot that wants to get out there and blow things up to free his people. That's what he does.
• The researcher that you bust in on in the caves? Shoot your way into an unprotected hidden research facility and see what the scientists inside do

I believe you're thinking more from a human point of view, or your own.

Now there are definitely other ways the AI could be improved, probably discussed here before, but AI that is super smart or realistic in every way is not generally desirable in game design.

One of the biggest examples is squads of hostiles that will wait outside doorways when clearly they could swarm into a room and surround you to take advantage of their numbers. That was intentionally as such to increase survivability and allow for extra tactics around doorways. (There are some other ways you can even exploit the enemy AI, but I won't point them out here because they're things you can figure out on your own )

Predictable not-entirely-smart AI mixed with numerous game mechanics (and multiple types of AI behaviors) is better than realism.

[Laida]

Spoiler (click to show/hide)

Shoot your way into an unprotected hidden research facility and see what the scientists inside do

Surprisingly, I actually worked in the something like research facility in real life for several months. It is funny to imagine it, but if somebody armed broke in to my research facility, I'd try to take my flash drive containing work data and flee to the safe place as soon as possible and then call for cops.  If I were more strong, I'd maybe tried to take out and keep safe one of neutron modulators, because it was the most valuable thing there as I know. But it is also really heavy, so I'm not sure about it. 

So, if we talk about tinkerer, he could flee to the Zion in this situation, or maybe just "slip back to the shadows"

There was once another sad situation about tinkerer, who tried to tell about code able to hack behemoth. He jumped around and round around me until enemies spotted and killed him.  According the plot, he was smart enough to find this code, but didn't manage that there is dangerous to walk and jump around so recklessly. This is... strange in my opinion 

I believe you're thinking more from a human point of view, or your own.

Yes, of course. Every time when people try to imagine an intelligent being, they actually imagine something with human-like mind, because we haven't examples of other type of intelligence in the real life. Modern deep learning and other "intelligence" artificial systems are still very-very far beyound self-awareness. And when we try to make some artificial intelligent agent, in the best case it have a deep imprint of that it's creator considers as "intelligent".

One of the biggest examples is squads of hostiles that will wait outside doorways when clearly they could swarm into a room and surround you to take advantage of their numbers. That was intentionally as such to increase survivability and allow for extra tactics around doorways.

Yep, I understood that this was made intentionally.

The thing about AI development is that it's an endless rabbit hole. I know because I've spent literally two entire years of my life developing smart AI for a game before. In the end it's not worth the vast amount of work involved and I decided I wouldn't expend lots of resources on diminishing returns like that. The time is far better spent adding other content and features.

Yes, AI development is really hard, that's why we very rarely see AI which is comparable with human player in games. But it is theoretically possibly. I imagine that it would be really cool to see good AI in the game about robot civilization, because it is so thematic there! But I know that this only one of many suggestions and possibilities how to improve the game. Players definitely want zilion cool features in games, just because they will not implement them by themselves. 

As a maybe little off-topic:

Spoiler (click to show/hide)

I'd like to try to develop game AI and to experiment with it by myself when I will have enough time, but it is near impossible with my current work and learning in post-graduate school at the same time 
AI in your game (as in 99% of games, I think) seems hard-coded. I'd like to experiment with genetic algorithms sometime, like in this videos, but maybe I will try to do it in more complicated envionment, than there:


Maybe you will find it funny too 

[Kyzrati]

Hehe, yeah genetic learning is awesome, something I've always wanted to experiment with ever since I learned about it, too. The thing is most game systems are hard to model well in a way that genetic AIs can take advantage of. Even GOAP is hard to model in games much simpler than Cogmind.

But in the end really smart AI is rarely worth the trouble, at least for most types of games. Cogmind uses basic FSMs like a lot of games, very good for control of the design, which is important in keeping outcomes generally predictable given a constant environment. Unpredictability is fun in certain games, but there are already plenty of sources of challenge in Cogmind that this doesn't need to be one of them

More content and features add a lot more value compared to further AI adjustments. In that vein, additional types of robots and factions with their own new AI just happen to be included on the list of stuff I'd like to do. No idea how far we'll get with that though, since I've gotta stop some time if support for development drops off too much... Time will tell!