Kael

Operator
  • Content count

    138
  • Joined

  • Last visited

About Kael

  • Rank
    Ranger School Dropout

Profile Information

  • Gender
    Male

Profile Fields

  • Platform
    PC gamer
  1. Epic rant video incoming

    Not just AMD/Nvidia, but Intel has a neural net processor and Apple/Google have rolled a few AI processors into their latest phones. All this equipment does is serve to provide the power to do more with AI, but it doesn't mean jack if the developers / engine-houses / publishers don't wish to put dev time into it. Single player games are falling to the wayside and the ones that do exist, have their established AI - the "good enough" thought. It's an afterthought. We're more likely to see development in AI learning to play humans in multiplayer (as that's fun for researchers to do) versus creating a more immersive experience in single player. It's just like bots of the past, why program bots when you can just tack on multiplayer? Why create elaborate decision trees, when you can just script it and funnel the player into the script (Call of Duty style)? It was painfully evident of the popular dev/pub opinion when the RainbowSix franchise was relaunched as an MP only game. Zoom is right when he says that TacShooters drove the AI development. Many games have super soldiers (the player) as a mechanic and just mask poor AI with more enemies, the thinking being that the short lifetime of an NPC will mask their stupidity. Digital whack a mole. The TacShooter was different since life was so limited, you could get one shot, and you could one shot the enemy. The tactical (almost puzzle) aspect of the game required more advanced techniques, as simple AI wouldn't cut it.
  2. How should AI react to Ghillie Suits?

    I don't think actual AI-actor shape recognition would be feasible computationally in games for quite some time (Zoomba can correct me if I'm wrong). We take for granted the amount of work our brains do giving us sight, for a computer to do it, it's not like you can just render a scene from the AI's perspective or hook up a camera, it has to be able to understand the data it's getting. That's why it's easier to do metadata queries, weighted values, thresholds, and logic trees (you can see this stuff in some of Zoomba's videos) to create the illusion of vision, since this is stuff the game engine can provide.
  3. Antibes map (WIP)

    Might want to look into Houdini '16. Seems to be pretty powerful for terrain/procedural generation. There's more videos under that account.
  4. My critique of the player controls

    Open world games come to mind (besides the others mentioned). I switch between a controller and KB/Mouse for games like GTA5/MGS5/Wildlands while playing. You need the full range of a controller's triggers/joystick to simulate pedals/steering correctly, but the fine precision of the mouse for firearms stuff.
  5. Just put a smartphone in their hand and the obliviousness to everything around them can be a realistic feature.
  6. AI work this week

    As others pointed out, this isn't the place for this feedback. Zoombapup is the AI specialist. Telling him to work on network code is like telling your plumber to work on your car engine. Sure, he might be mechanically inclined, but he's going to take a helluva lot more time fixing something as it's not his area of expertise.
  7. Extent of tactical items?

    Guess it depends on how realistic they want it to be. Lockpicking is good for stealth entry. Games generally do it in one of two ways, either the meter/timer based hold the button way (R6 did this) or the mini-game (Splinter Cell did this). The other options should serve a purpose. A wedge/bar/ram make sense for a dynamic entry where you don't want to destroy something on the other side (breaching charge). Could fit in-between. Stealth: Lockpicking Fast Entry, non-stealth: Wedge/Crowbar/etc. Dynamic [Loud] Entry: Explosives/shotgun
  8. Road Editor - looks pretty sweet

    That's pretty cool. Sorta like Cities Skylines. I'm guessing it needs to be supported by the base game because maybe they're using math to generate the roads and it needs a translator? You'd think they'd include a way to just export the road structure as objects, so they'd be supported outside without a base-code inclusion. I kinda wish Epic would put together a proper map editor that was designed for larger maps (the current one is sorta leaned towards UT style). Not sure if you've seen Planet Coaster's official release, but despite being a park management game, it has one helluva level editor built in. Everything from terrain / textures / lighting / sound / assets is easily implemented by the user. Something like that with maybe Cities Skylines style procedural generations (for fences, power lines, roads, and repeated assets) would be a killer toolset. The level comes alive in the small details, so being able to work quicker allows the mapper to spend more time refining/polishing. I occasionally check out cool level design videos and if you're looking for inspiration/just-to-look, I'd recommend checking out the Arrowhead Junction series on YouTube. It's a modded Cities Skylines and the level of detail they use is awesome, I sometimes get tricked thinking it's a real photo because they're basing it off of an area I've driven through many times.
  9. Gatehouse (map WIP)

    Looking very good. Especially given the tools you have to work with.
  10. Is it time to start on squadmates?

    That's a pretty significant upgrade. About time they addressed that. I wonder if you can stream the bakes mid-match (e.g. Triggerable)? That'd allow for power cuts and light switches. Definitely adds a more tactical element to mission design.
  11. AI work-in-progress update video #12

    And that could have been adjusted for with some variation in spawn locations. I can't stand when level designers put one spawn location for the enemy troops. If it had to select randomly from different locations, it'll help mask the spawn in. On top of that, some level designers could lazy and put the spawns out in the open, so they're sometimes literally noticeable. I think John and co will do it right and make it so it's not noticeable or they're spawned in at the beginning.
  12. AI work-in-progress update video #12

    It's about giving the option to the mission designer as well. There's times when you can do a proper spawn in of AI without it being obvious. The way CoD does it is obvious. You could put them in an inaccessible barrack that only open (become accessible) when an alarm is triggered, spawning them in right before the door unlocks. Or so they don't spawn into these locked rooms until a player reaches a certain distance, simulates a sleeping AI tango without having them actually taking up CPU cycles. You could have listening events that simulate a threshold for them to spawn (e.g. you're too loud). Or use it for reinforcements to spawn at the edge of the map, so you don't have to literally have 30 AI guys standing around in a non-accessible (to player) area. I think it's a fine idea as long as they spawn far enough from the player so they have to traverse the level somewhat to get to the action, adding an unpredictability to their approach. I can't stand COD's wave of spawns around the corner, removes the realism and makes for a constant "there will always been an enemy there" play through. Just like R6 Vegas did. If done right, it's a good tool to have. Personally, I'd rather have the CPU cycles used for smarter AI versus having every AI spawned in the map at once but dumb as heck.
  13. Wow, excellent addition. I kinda feel like a lot of games just take the easy way out and make it so bullets don't hurt each other.
  14. Pankisi Gorge map (WIP)

    I don't think the GB team know how they're going to handle user created maps yet. It's kinda a post release thing. But I could be wrong. You can do the push method, where a new map gets pushed to the player on a first load (longer). Or the reject method, where you get rejected by the server and given a link to acquire the map. A more advanced method and because assets can grow in size (textures, etc.), there might need to be a way for there to be a community asset pack that gets updated periodically. That way maps that use the same assets load quicker because they already exist. ARMA (as example) has always had community packs that act as frameworks for mods, so it cuts down on redundancy. That would require a check to see if an asset exists on the client and download if necessary. So a user who has a lot of user created maps would acquire them quicker than someone who doesn't, as their asset storage would need to acquire more objects.
  15. Animations feedback

    Did you just turn Jamaican or have a stroke?