Human Contact

March 31, 2010 | Filed under: game design

I spent today mostly working on campaign stuff. One of the features in the campaign game is that player-designed fleet deployments show up as rival fleets in the campaign. So rather than just facing a fleet designed by me, and always the same, in a given battle, it’s a fleet someone else designed. Another GSB player.
As I play-test this, I find it really cool. I have no idea what each enemy fleet will look like until they appear, and it seems somehow ‘more exciting’ than the idea that its AI-designed, or a fixed fleet designed by me.

Why is this? And is this just me?

Playing against other people is often described by PC gamers as way better than playing against the AI. When you ask people why this is, there are a lot of reasonable explanations about humans being better at adaptive strategy, more creative, and with the ability to actually suprise or confuse the player. Certainly this is true of most RTS games and FPS opponents, at least currently.

I do wonder though, on how we will feel in five or ten years time when AI is that much better. There is nothing to stop AI analyzing every RTS or FPS players movements, and ‘evolving’ play strategies that mimic real human players much more accurately. In twenty years time do we really think that we will be able to tell if our RTS opponent is AI or Human? Assuming we don’t, will we *still* get an inexplicable buzz of excitement knowing it’s another squishy human rather than some clever AI?

I really think we will. There is something very primitivly rewarding about pitting yourself against (or co-operating with) a fellow human. We can’t speculate much about the motives or emotions of some AI, but the thought that there is someone, somewhere who just got their fleet pulverised by ours is just somehow much more rewarding. Blowing up AI pirates in eve is kinda fun, but blowing up other players who then have LOST their ship, is awesome, albeit a bit cruel. What do you think? Do you think it makes a difference if you are directly, or indirectly (think spore) playing against real human beings?

10 Responses to “Human Contact”

  1. Alehkhs says:

    Wow, I love the idea of player-designed content (fleets) showing up as my enemies!

  2. Fargo says:

    If nothing else the variety sounds like a great thing. Stats on how well your fleet design performed could be fun, like an indirect leader board. Might be a bit much effort for the value though, from your end.

  3. CountVlad says:

    I think the main difference between AI and humans is that AI is entirely logical, even if it’s told to do something randomly depending on a variable. It still has a ruleset that it must obey under all circumstances. Humans, on the other hand, may be entirely illogical in their approach. For example, they may lay out their fleet so it initially looks like a cartoon pair of breasts. An AI is not going to do something stupid like that unless it has been told to previously. It has no logical purpose.

    People, whether we know it or not, can work out what someone else is likely to do next based on their previous experiences with that person. We do it all the time without even noticing. AI usually reacts to user input, and not to what it thinks you will do next. That is why humans are more fun (and harder) to play against than a computer.

  4. Alex Vostrov says:

    I think that it matters.

    Chess AI plays better than humans now, but it’s still more fun to play against people. There’s always the possibility of making mistakes, amazing comebacks.

    Playing chess against AI is as exciting as wrestling against a bulldozer.

  5. Benji says:

    “I do wonder though, on how we will feel in five or ten years time when AI is that much better.”

    Has AI really gotten any better in the last 10 years? I feel like I’m seeing less examples of good ai than I used to see, now that games focus so much on presentation.

  6. Alex McLarty says:

    People. You can break them.

  7. One of the great problems in game-adversary AI is making it realistically DUMB. Playing Go is a hard problem, but playing a FPS or RTS game is a quite easy one, for modern hardware. Bots can never miss in FPSes, and can have 10,000 actions per second and perfect situational awareness in RTSes. Such enemies are obviously bots, of course, and nobody wants to play against them, but that’s not because they’re not “creative”.

    I think GSB could be an excellent candidate for an evolutionary AI experiment. If the game can be made scriptable enough, someone could make an AI that literally starts from the ground floor putting random modules in random ships, pits them against these neat-o player-made fleets in max-speed battles (preferably with graphics/sound turned off, for fastest performance), examines the results, mutates and evolves.

    You probably don’t want to do this yourself, since it’s not at all likely to be a nice little earner. But it’d be great if some other hacker did!

  8. cliffski says:

    Indeed. there was a point in the games development where I started doing a mutating AI that did exactly that, but I never got far enoguh for it to be usable, and I just ran out of steam at that point. but I agree that it would be quite a cool experiment. It would also be interesting to have AI that pieces together the most succesful ship designs and deployments that people upload to create even better ones as original, but player-influenced designs.

  9. Brit says:

    > “I feel like I’m seeing less examples of good ai than I used to see, now that games focus so much on presentation.”

    I wonder if that’s because there is less work going into AI (as you say), or if it’s because games are more complex – and, therefore, more difficult to program with a decent AI. Afterall, when a game has fewer options (as games in the past did), then it’s easier to program a decent AI. Today, there’s a whole bunch of options (because new features are a selling point), but it’s a whole lot harder to make an AI to balance those options against each other and pick the best. For example, it’s a whole lot easier to program an AI that can play tic-tac-toe than it is to program an AI that can play chess.

  10. Drewoid13 says:

    I like it! In GSB playing vs. another person is better than most other games. Because player designed fleets can only be so good, and the deployments the same, there is no fear factor like in Starcraft where the new player will get laced by a slightly more experienced one. (actually this is the reason I hadly ever play online mp).

    And while AI has the potential to be very good, intelligent and responsive, I find very few games that come close to even being half that good. HoI3 is one of the best so far whereas ETW is like trying to play chess with my goldfish.