To celebrate 10 years of AI and Games, there was one topic that warranted being revisited one last time, the AI of FEAR. But this time I wanted to do something special. Over the past 3 months I've sat down and had several (online) interviews with Dr Jeff Orkin: the AI programmer on FEAR. In this retrospective, we get into the backstory of how the AI of FEAR came to be. The academic influences, the crazy risks being taken, the lack of oversight, a pivot from 60s espionage to sci-fi horror, the creativity of the development team behind it, and the legacy the game has had in the game AI community. A huge thank you to Jeff for participating in these interviews. I've had a great time listening to his stories. Critically for me, FEAR was massively influential in my own career path and without it, my trajectory through grad school would have been drastically different. And of course, given it was the inaugural episode of AI and Games, the UA-cam channel may never have been created.
The idea that the enemies always talk to each other rather themselves is the genius of FEAR (saying “where is he?” “Behind the box” rather than “I can’t see him”)
People undervalue AI barking so so much. The more the AI shows the complexity of its decision process through gameplay elements, the more engagement the player will get out of it.
It's such a bizarre thing when you take it out of context, but players often think NPCs are 'smarter' when they simply say out loud what they're doing.
@@AIandGames I recently replayed F.E.A.R and I find myself thinking how game designers are a lot smarter when I hear 'Flush him out!' followed by *Clink Clonk* tipping me off I'm about to be naded than seeing a red UI indicator pointing to the naughty thing that landed next to me. One thing provides player with the information needed in a immersive and believable way, the other breaks the 4th wall reminding you how you are in a videogame and isn't expected to actually pay attention to any of the graphics but instead look and play that interface. 'AI communication' isn't just about what they're saying it's also so immersive because it provides directionality and opportunity to gain information about what occurs out of sight in a natural way. In a lesser game this would be interface indicators with action icons tipping you off how that purple health-bar hiding behind that box is about to perform some specific action.
@@AIandGames I guess its expectations of people of what NPCs can do. Game needs to show to the player that it is more complex or player will expect it to be simpler by default. Although complex dialogue systems will need to be backed by what NPC does in game world or illusion will fall apart really quickly. That was my experience with Half-life 2 Combine soldiers, they are much simpler than their speach system suggested at first.
This game really shines when you make a no-slowmo playthrough, and the higher the difficulty the better. That's how you really get to appreciate the smart AI, instead of plowing through them like Neo, leaving them no time to do much of anything.
Def. Recently replayed it for the first time since it's release days that way specifically to study the AI and level design. Unless you dial the difficulty up and use your own reaction times you're unlikely to ever notice just how well this is made. You need that respect not to risk exposing yourself to the AI's sight-lines to give it space to do it's thing. Unless there is some symmetry in lethality clever flanking behavior just isn't impressive in games, it requires a need to outmaneuver for survival to take center stage. Recently played Division 2 and juxtaposed with F.E.A.R observed how the AI occasionally did things that could've been impressive but in that game it doesn't matter one bit because their (likely accidental) clever movement is spoiled by the interface and being able to tank a lot of damage makes it pretty pointless even if they succeed out-flanking a player. Made me wonder how many games with elaborate behavior trees since F.E.A.R are just like that in wasting the effort thru dumbed down interface and overly forgiving gameplay. If we played F.E.A.R with a permanent radar and near wallhack vision where everyone where located it wouldn't been remembered for much of anything.
Honestly, though my current playthrough is normal w/ Slowmo (it's the console port, I'm doing normal to get used to the controls), my preferred way to play F.E.A.R., its expansions and F.E.A.R. 2: Project Origin is Hard w/ Slowmo. On hard, because the enemies are so skilled, slowmo is basically your trump card.
Thank god I was not the only one that though that way I played it has kid and I only found out about how to use the slow motion AFER finish the game and I'm grateful I never did because not only made every action more tactical but also keep me more invested on the story I felt like any mistake would actually have consequences which tied with the story itself were a game is actually trying to kill me rather than teach me were to go but lore wise I still think the slow motion was actually a good implementation since it helps the players that didn't have that much skill to actually progress, but I still wished the slow motion was shorter or more balanced so that it wouldn't get that much abused wile still offering that window of help to the players, Nonetheless FEAR remains 1 of my favorite games till this day along other horror shooters like dead space, condemned: Criminal Origins and so on, those dammed horror tapes of Alma still hunt me to this day...
On your question as to why FEAR doesn't seem to have been surpassed 20 years later i think it's a two sided issue. When you look at the top games you often reference too it's FEAR ( G.O.A.P ), Half-life ( finite state machines ), Halo ( behavior trees ). But one thing all these games have incommon are aggressive enemies with varied behaviors and they communicate ( bark ) a lot. I think this is also what Jeff also points out where his close working relationship with animators, level designers and audio recording all involved him to some degree. The AI is extremely good but the player wont know if it's not communicated to them by the AI putting the spotlight on itself ( sort of that stage for the NPC quote he mentioned ). I think it's understimated how much impact this has. Then the other side of the issue is like he points out how risk adverse game development has become. Combine that with a large part of the industry now being formally educated in the use of ready made tools like unreal, unity etc instead of talented programmers with a propensity for games . And it seems the chances for something interesting to happen are slim to none. I'd fully expect there to be more impressive systems out there by now and maybe there are but they aren't communicated well. It seems there isn't much interest in making smarter AI oponents in the space where budgets would allow such investments and i suspect if anything will happen it's going to come from the indie space.
F.E.A.R. is just hands down one of the best shooters of all time. Genuinely the best squad AI in a game ever. UT2004's bots come a close 2nd place but for tactical combat the Replicas win. Hearing Orkin's story was amazing, what a super talented genuine guy! Great and well edited video!
Replaying this game right now and it holds up so well… the bits that haven’t aged are the cheesy horror and the story but gameplay-wise it is just 100% all the way. Nothing else like it
I’ve been using your videos as the core material for introducing my students how fascinating GOAP in their project. The feedback from my students has been overwhelmingly positive-filled with excitement for the simplicity yet surprisingly incredible performance. Will definitely continue to inspire young minds with the same sense of wonder that F.E.A.R.'s AI instilled in us, the feeling when we were first ambushed by those NPC clone soldiers.
The fact that Fear AI can intentionally drive you back into choke points so that they can lower their own risk and grendade an opening instead is insane.
Fear should definitely get a remaster to fix the bugs and improve the visual fidelity but the gameplay and story is already perfect! I'm so glad this channel exists and I hope you come out with more banger content as AI becomes more advanced.
i wish AA quality games like FEAR were still made today. Sometimes a hyper polished, focused, masterfully made yet smaller-in-scope game is all I want.
Fear might be AA by today's standards but it's actually AAA. When it came out it blew people away and was and still is a quality shooter and story experience.
My favourite chatter from the enemies was "Where should I go?" "Shut the fuck up!" It really got across the stress of the situation. "Where is he? Fuck! Oh shit! He's wiped out the whole squad!"
Commenting for the algorithm primarily, but also to just give my two cents on the final big question. I think a big part of what makes the A.I. in F.E.A.R. still hold up so well is the way in which it was so closely integrated with the level and audio design, animation work, and the gameplay in general. The enemies feel interactive, reactive, and exciting because so much of what they do relates to their surroundings, as well as to the presence and their knowledge about the player character. The use of calls and responses in their dialogue, and actions like firing while retreating, diving through windows, and hiding make them seem especially dynamic, even if most of those are ultimately achieved through relatively simple audio cues and pre-built map pieces being triggered by the A.I. Also, it may seem like a strictly stylistic gameplay choice, but the ability to slow down the action and see combat unfold more clearly also helps to highlight the quality of the incorporation of the A.I. further. All in all, while I do think it's fair to argue that more recent games using GOAP or other systems have more advanced A.I., it's the close integration of the enemy A.I. with F.E.A.R.'s various systems that make it so impressive. And with how segmented and often disparate various parts of game development typically are in the modern industry due to larger teams and longer development times, achieving such a tight connection between A.I., gameplay design, level design, and presentation simply isn't going to happen as easily. F.E.A.R. came at just the right time.
46:45 In my view as someone who is currently attempting to build a interesting AI system for a tactical shooter type game I think what's 'haunting the industry' here is how much knowledge is necessary to have rolled up in those few people with real influence of how to allocate the time and budget to see this thru. To know how to craft gameplay like the beloved parts of F.E.A.R you need a high level understanding of how to design the technical state machines, behavior trees, path-finding that is mirrored by the spatial design of the level by someone who is also a proficient gamer who understands how to hold angles and what affordances different cover and placement will provide. The difficulty is in how to create a scenario that forces the player to traverse the environment mindful of exposure and have the AI functionally play 'cat and mouse' attempting to get the drop on one another. Games of this era was still made by teams of all enthusiasts who built games they themselves would want to play. A small team that can pivot on a dime where everyone understands a lot about the thing they're attempting to build at an enthusiast level are a lot better poised to get it right than a gargantuan team where you have to clear every decision with people who may or may not have the slightest clue what you're talking about or even able to play your own game well. F.E.A.R itself is very start/stop with those ingenious 'combat space arenas' we all remember being dotted thruout a game intermingled with long segments of the same sort of linear corridor geometry guilty of all the same mistakes how forgettable encounters in titles before/after tend to play out. Replayability needs unpredictability and that is something even F.E.A.R only succeeds with in a handful of places where the start configuration of the encounter doesn't determine how the rest unfolds. That slice of F.E.A.R that get all of this so right is what needs to be lifted out and studied in great detail, it's a synergy where ones understanding of player tactics, AI tactics and spatial level design all play equal importance.
What an excellent revisit/retro celebration, hard to believe it's 20 years ago already! Started with Shogo and although early 2000s Monolith may be a niche it definitely was mine - would be wonderful seeing NOLF3 someday with all the advancements and new discoveries Jeff make on dialogue/etc, including what comes out of the recent language model chatGPT hype. I read his paper back in uni days, the time he spent in the office back then have definitely paid off. FEAR is the mind thriller!
@@AIandGames Hey, thanks for the recommendation! I really love FEAR (and FEAR 2 is also OK) and I played it and its expansions just last year. I'm currently playing TimeShift, it's obviously not as good as FEAR, but it has some neat slow-mo and time warping tricks in it, it's overall pretty good. But I'm definitely giving Trepang2 a try 🙂
@@dodolurker Yeah FEAR 2 is okay, didn't really enjoy FEAR 3 personally (different dev team too). I've never played TimeShift actually. Just noticed it's like £5 on Steam, so might give that a shot. But yeah Trepang2 is 'what if FEAR in 2023'. It's pretty neat.
@@AIandGames TimeShift is... I think one reviewer summed it up pretty well - it looks like a FEAR expansion pack and really wants to be Half-Life 2 😀. Whereas those games are pretty much timeless, TimeShift feels like a shooter from the era. It's also pretty much forgotten today. The only reason I started to play it is because I bought it when it was originally released, caught a glimpse of the box on my shelf and thought "hey, why not give it a go, I didn't finish it back then". But yeah, it's definitely not bad so far, if you can get it on sale, check it out. I also understand that the GOG version of TimeShift is more patched than the Steam version. It's currently on sale on GOG as well, €9 right now. It also needs another patch, otherwise it crashes when you try to run it, I think it's something to do with too many cores on today's CPUs. But the patched exe is easily found online 🙂
6:01 Ah man, this brings me back to my memory lanes. I was too young to combat the jank and couldn't pass much further than the start, but man I was trying enthusiastically. Good times.
Great video. I've always has great respect for Jeff Orkin and have tried to implement his GOAP stuff multiple times (out of interest mostly, I'm not a game dev, just love the stuff) so it's really great to see him and hear him talk about this. Loved it.
Could see yourself climbing a ladder or running around.... or flying scissor kicking people people in the head in slow mo. How was there never another game like this one. Pure joy.
In Montezuma's Return you could also see your body when looking down. Not sure if it was the first, but it predated Trespasser for a bit. A much more niche -and weird- game though (and the English version was released later).
Monolith's 'Blood' had an ad with a guy in a bathtub full of blood, holding two pistols I think, with the caption "Blood. You're soaking in it." A tongue in cheek take on the Palmolive soap ads of the time.
I remember a lot of FEAR being held up as a high watermark for AI in video games, still is to this day like in this video, and it was one. But I remember a somewhat hilarious moment playing it when I would kill an enemy, close a door, another enemy would open the door, I'd kill them, and I quickly had a mound of enemies piled up right at the door that the other soldiers kept running into and and instead of reacting in any sensible way they just kept coming though and dying lol. A moment of artificial dumbness. Also made me paranoid of climbing game ladders from the jump scares I'd still love to see a new modern FEAR game.
I think this is the best way to celebrate the 10 years, and I love all the context that Dr Orkin gives us. He really walks you through how GOAP came to be, and I loved that. Amazing, simply amazing video as always!
Sorry to go a bit off-topic but Alien vs Predator 2 was such a good game. I dunno but it feels like it didn’t deserve the acclaim it should halve. Both the amazing single player story and the asymmetrical multi player where great games. Got some great memories of the latter with death matches.
I think the first game I remember showing the player-character's full body in first-person was Operation Flashpoint (ArmA: Cold War Assault). I think the ArmA series would make a very interesting video for this channel - the AI had to be able to do their job wherever they were on these huge maps for the time - react to threats of different types, potentially coming from any direction, with multiple units maintaining cohesion
F.E.A.R.'s AI is still unrivaled in the artform. The Replicas and Armacham forces are so much fun to fight against because the AI was designed to be as smart as possible, thus turning who would normally be generic mooks into dangerous threats whom you have to outsmart. And one of the reasons why I'm in the camp that also loves F.E.A.R. 2 is that the GOAP AI made a triumphant return, and is still fun to fight against in that game.
That was a really great watch, thanks 🙂. FEAR is definitely still awesome today. It would be very cool if you could track down Tom Leonard, formerly of Looking Glass Studios... I'm still awestruck about his work on Thief: The Dark Project 🙂
I'd love to do something on Thief at some point. I should go and drop Tom a line, he's worked on all sorts of interesting stuff over the years (Portal, Half-Life 2, Left 4 Dead).
Something I learned from this video is that today’s gaming is missing risk takers. The fact they was using all their resources just to make the AI challenging is truly amazing awhile everyone doubted monolith decisions they excelled by making a timeless masterpiece
Its cool for someone like me to hear about all the work that went into this game when all I can remember is how scared I got almost 20 years ago when you go down that ladder and you see that girl where you just were
Fascinating video! I've always wanted to get into game dev, and similar to Jeff, I was interested in animation and graphics part but AI and behaviour programming sounds so much more interesting to me.
excellent video and interview. you mention around 21:15 that quake 3 was the first to solve or popularize or standardize navigation meshes but I am curious to hear about original unreal's method of solving this problem or if they also used navigation mesh at that time. unreal's bot support was the king for quite a while and even compared to quake 3 has its own qualities. it's possible these bot paths were all hand crafted but I feel like I remember playing custom maps with them. I know nothing and any information regarding the topic would be very cool to learn about
So Quake 3's AAS system isn't the first solution for navigation, but it's largely the one that would later be standardised into other games. I haven't covered the original Unreal (yet), but I did cover the the navigation system used in Goldeneye 007 a couple years ago which has series of (manually placed) nodes in the world that characters could move between.
Well, thanks to Jeff for making an AI that is fun to setup in combat arenas with the SDK, altho having to manually create the navmesh is something x) Also amazing video!!
The fact that Monolith dipped their toes in fps games as varied in tone as FEAR and NOLF....is amazing. Such wild swings are practically unheard of in today's times. Companies today are shit scared to dip their toes into anything remotely "wild" and experimental....least they rock the stockholders boat. AAA gaming today is utterly bland and soul less.....and it doesn't look like it's going to change anytime soon. At least we can go back to these wild games that scratch that itch.
A.I. drives engagement in games, While graphics have been steadily improving A.I. seems to change in leaps. Some studios just find out what is already out there and is the bare minimum to get the game "working". I hope the future holds some great experiences. As a hobby game programmer when I was younger without knowing what it even was I made a rudimentary Finite State machine to control my CPU enemy.
FEAR is known for its amazing AI, and still being one of the most challenging shooters of all time, but. I remember it for the melee combat system. Especially in multiplayer deathmatch, it was absolutely disgusting. You ran extra fast, and your flying and sliding kicks made it so that you were harder to get shot and they one-hit killed. There's a keybinding to holster your weapon that the game never tells you about. So you can throw down any time in the campaign as well.
You can change the runspeed when you create an MP game, but yea most servers have it really high. But I liked it, the slomo TDM is the most satisfying thing that I could play anyday
@@AIandGames oh sweet! The SDK should be easy to get a hold of and the code is written mainly in good ol UScript. Thank you for all that you share on here, love your work, mate!
The video is great, awesome insight and i love how he makes it seems simpler than it really is. Just one observation the sound in the background while Jeff is talking makes quite hard to concentrate on the subject... had to re-watch the video because the background noise is too distracting
Great video but I had to skip the chapter “Orkin’s path into games’ due to the irritating background noise throughout. EDIT: Seems like it persists in some form for the entire video. What a shame.
Awesome video, man! This has to be one of the greatest retrospectives I've seen on FEAR! Now, I would like to give you some feedback regarding the Brazilian Portuguese translation of the video's subtitles and description. I don't know if you got someone to do it or if you used AI/Machine Translation, but the amount of mistakes and overall lack of care are really concerning. I'm not sure about the quality of the other translations, but the PT-BR one is just not acceptable. In regards to the subtitles, they are too literal and badly formatted for a good reading experience.
I never played F.E.A.R. before, i know i watched videos the AI is very smart, so thanks to this video i bought the game and holy crap, community weren't liying with the AI of this game, a freaking enemy run around a building just to take me by surprise and attack me.
I would like to hear what he would do different now, if he had all the modern benefits of developing a game, that he mentioned and was making a game like FEAR again.
Unless they patched it, i remember that the loading of the soldiers in an area would slow down the game a bit and all you had to do to clear an area was walk through a door to trigger the enemies and then they would all funnel through the door in a line.
I would be very interested in your thoughts about the fact that most of the games we recognize as having impactful AI were released so long ago? AI breathes life into games, but these days people seem more interested in using AI to generate silly pictures… Will it come back around?
AI is still improving, its just that the advancements are iterative and less impressive. these are where the technology that have been improved on started. With the exception of racing games which use machine learning algorithms for there AI.
This was a great interview! It is sad seeing a lot of AI in games being just kinda of bland and good enough. Disappointing to hear he's doing genAI now. Hopefully it's using ethical training data.
We'll be discussing some of the stuff Jeff's doing now in a future video, but I've seen the tech demo he's built and it's really cool. It's a rather smart application of generative AI, in that it only uses it as part of larger system that still relies on existing approaches (including planning). And critically, the system relies on users providing their own training data, rather than stealing it from the internet.
If you can find a physical copy it's backwards compatible on Xbox One and Series S/X. For context: FEAR 2 and 3 were published by WB after they bought up Monolith in 2004. But the original was published by Sierra, and it'll be some rights/royalties issue I suspect between the two that prevents the digital listing on console.
2025 The game begins with a call to the senator. A female voice reports problems with the Source: “Riot. Fettel has taken over control of the prototypes." Armacham Technology was a weapons developer with a military program: an army of clone troopers subordinate to a commander who controlled them using telepathy. Paxton Fettel uses his telepathic abilities to take control of a battalion of clone troopers. Fortunately, to stop an armed conflict, it is enough to eliminate the commander, after which the clones will cease to act and will be absolutely safe. And finding Fettel won’t be difficult: he has a sensor implanted in his head that gives away his location. He will not have any protection, one of the conditions of the program is the safety of the commander, he does not fight along with the rest of the soldiers, but only controls them at a considerable distance. The protagonist of the game is a new operative officer “F. E.A.R.”
Damn it. We would have had Nolf 3. My favourite Monolith series. This sucks. I don't know why companies look at other companies and decide this won't work even though lot of times a good unique game succeeds. Nolf 2 actually have really good AI. There is an article from Monolith where GOAP is explained with Nolf 2 as base. It is quite clear that Nolf 2 is the start of Fear AI.
None of the developers who worked on fear are there anymore. Do you trust a bunch of new people to have the same passion for the project, or would they just see it as a product to generate money
Sorry, it looks good and I'd love to hear the interview, but I can't be in the same room with all the action going on around the borders of the interview video screen. Vertigo and motion-sickness inducing, turned up to 11. Obviously you can't redo this video, but maybe keep it in mind for future ones.
Noted! Thanks for the feedback (and sorry, not something that I thought of when editing it together). There is the *written* version over on the AI and Games website if you like, y'know... words
As a single-player experience, I'd say FEAR is just about the FPS made since... 2005? It always makes me sad to listen to people like Jeff Orkin speak about game dev in the 1990s and early 2000s. These guys were always trying to experiment and make cool stuff. Of course they had to worry about sales, but it wasn't like today where every single facet of most games gets degraded in order that the dev can monetize the "enhancements" and "fixes". It just doesn't feel like devs have a passion for what they make, nor any concern for the consumer's experience.
To celebrate 10 years of AI and Games, there was one topic that warranted being revisited one last time, the AI of FEAR. But this time I wanted to do something special. Over the past 3 months I've sat down and had several (online) interviews with Dr Jeff Orkin: the AI programmer on FEAR.
In this retrospective, we get into the backstory of how the AI of FEAR came to be. The academic influences, the crazy risks being taken, the lack of oversight, a pivot from 60s espionage to sci-fi horror, the creativity of the development team behind it, and the legacy the game has had in the game AI community.
A huge thank you to Jeff for participating in these interviews. I've had a great time listening to his stories. Critically for me, FEAR was massively influential in my own career path and without it, my trajectory through grad school would have been drastically different. And of course, given it was the inaugural episode of AI and Games, the UA-cam channel may never have been created.
Ok, let's pump that algorithm because F.E.A.R. is a masterpiece and so is this video!
Damn right its a masterpiece!
YES!
🎉🎉🎉😊
A remaster would be so welcome in gaming
FEAR on impossible difficulty without using time slow = best single FPS experience.
The idea that the enemies always talk to each other rather themselves is the genius of FEAR (saying “where is he?” “Behind the box” rather than “I can’t see him”)
People undervalue AI barking so so much.
The more the AI shows the complexity of its decision process through gameplay elements, the more engagement the player will get out of it.
It's such a bizarre thing when you take it out of context, but players often think NPCs are 'smarter' when they simply say out loud what they're doing.
@@AIandGames I recently replayed F.E.A.R and I find myself thinking how game designers are a lot smarter when I hear 'Flush him out!' followed by *Clink Clonk* tipping me off I'm about to be naded than seeing a red UI indicator pointing to the naughty thing that landed next to me. One thing provides player with the information needed in a immersive and believable way, the other breaks the 4th wall reminding you how you are in a videogame and isn't expected to actually pay attention to any of the graphics but instead look and play that interface.
'AI communication' isn't just about what they're saying it's also so immersive because it provides directionality and opportunity to gain information about what occurs out of sight in a natural way. In a lesser game this would be interface indicators with action icons tipping you off how that purple health-bar hiding behind that box is about to perform some specific action.
@@AIandGames I guess its expectations of people of what NPCs can do. Game needs to show to the player that it is more complex or player will expect it to be simpler by default. Although complex dialogue systems will need to be backed by what NPC does in game world or illusion will fall apart really quickly.
That was my experience with Half-life 2 Combine soldiers, they are much simpler than their speach system suggested at first.
This game really shines when you make a no-slowmo playthrough, and the higher the difficulty the better. That's how you really get to appreciate the smart AI, instead of plowing through them like Neo, leaving them no time to do much of anything.
Def. Recently replayed it for the first time since it's release days that way specifically to study the AI and level design. Unless you dial the difficulty up and use your own reaction times you're unlikely to ever notice just how well this is made. You need that respect not to risk exposing yourself to the AI's sight-lines to give it space to do it's thing.
Unless there is some symmetry in lethality clever flanking behavior just isn't impressive in games, it requires a need to outmaneuver for survival to take center stage.
Recently played Division 2 and juxtaposed with F.E.A.R observed how the AI occasionally did things that could've been impressive but in that game it doesn't matter one bit because their (likely accidental) clever movement is spoiled by the interface and being able to tank a lot of damage makes it pretty pointless even if they succeed out-flanking a player.
Made me wonder how many games with elaborate behavior trees since F.E.A.R are just like that in wasting the effort thru dumbed down interface and overly forgiving gameplay.
If we played F.E.A.R with a permanent radar and near wallhack vision where everyone where located it wouldn't been remembered for much of anything.
Honestly, though my current playthrough is normal w/ Slowmo (it's the console port, I'm doing normal to get used to the controls), my preferred way to play F.E.A.R., its expansions and F.E.A.R. 2: Project Origin is Hard w/ Slowmo. On hard, because the enemies are so skilled, slowmo is basically your trump card.
Thank god I was not the only one that though that way I played it has kid and I only found out about how to use the slow motion AFER finish the game and I'm grateful I never did because not only made every action more tactical but also keep me more invested on the story I felt like any mistake would actually have consequences which tied with the story itself were a game is actually trying to kill me rather than teach me were to go but lore wise I still think the slow motion was actually a good implementation since it helps the players that didn't have that much skill to actually progress, but I still wished the slow motion was shorter or more balanced so that it wouldn't get that much abused wile still offering that window of help to the players, Nonetheless FEAR remains 1 of my favorite games till this day along other horror shooters like dead space, condemned: Criminal Origins and so on, those dammed horror tapes of Alma still hunt me to this day...
On your question as to why FEAR doesn't seem to have been surpassed 20 years later i think it's a two sided issue. When you look at the top games you often reference too it's FEAR ( G.O.A.P ), Half-life ( finite state machines ), Halo ( behavior trees ). But one thing all these games have incommon are aggressive enemies with varied behaviors and they communicate ( bark ) a lot. I think this is also what Jeff also points out where his close working relationship with animators, level designers and audio recording all involved him to some degree. The AI is extremely good but the player wont know if it's not communicated to them by the AI putting the spotlight on itself ( sort of that stage for the NPC quote he mentioned ). I think it's understimated how much impact this has.
Then the other side of the issue is like he points out how risk adverse game development has become. Combine that with a large part of the industry now being formally educated in the use of ready made tools like unreal, unity etc instead of talented programmers with a propensity for games . And it seems the chances for something interesting to happen are slim to none. I'd fully expect there to be more impressive systems out there by now and maybe there are but they aren't communicated well. It seems there isn't much interest in making smarter AI oponents in the space where budgets would allow such investments and i suspect if anything will happen it's going to come from the indie space.
I think that's a fairly good assessment of the situation.
Revive this into a franchise NOW!
F.E.A.R. is just hands down one of the best shooters of all time. Genuinely the best squad AI in a game ever. UT2004's bots come a close 2nd place but for tactical combat the Replicas win.
Hearing Orkin's story was amazing, what a super talented genuine guy!
Great and well edited video!
Half Life 2’s AI is fantastic, but the level design really didn’t help it shine.
Replaying this game right now and it holds up so well… the bits that haven’t aged are the cheesy horror and the story but gameplay-wise it is just 100% all the way. Nothing else like it
Engagement! Comments! F.E.A.R.! Games and AI!!!
I’ve been using your videos as the core material for introducing my students how fascinating GOAP in their project. The feedback from my students has been overwhelmingly positive-filled with excitement for the simplicity yet surprisingly incredible performance.
Will definitely continue to inspire young minds with the same sense of wonder that F.E.A.R.'s AI instilled in us, the feeling when we were first ambushed by those NPC clone soldiers.
The fact that Fear AI can intentionally drive you back into choke points so that they can lower their own risk and grendade an opening instead is insane.
Oh fuck yes! Redo of F.E.A.R.
Fear should definitely get a remaster to fix the bugs and improve the visual fidelity but the gameplay and story is already perfect! I'm so glad this channel exists and I hope you come out with more banger content as AI becomes more advanced.
i wish AA quality games like FEAR were still made today. Sometimes a hyper polished, focused, masterfully made yet smaller-in-scope game is all I want.
There are, theres games like Selaco and Trepang 2 which are high quality games inspired by FEAR.
Fear might be AA by today's standards but it's actually AAA. When it came out it blew people away and was and still is a quality shooter and story experience.
My favourite chatter from the enemies was
"Where should I go?"
"Shut the fuck up!"
It really got across the stress of the situation.
"Where is he? Fuck! Oh shit! He's wiped out the whole squad!"
Commenting for the algorithm primarily, but also to just give my two cents on the final big question.
I think a big part of what makes the A.I. in F.E.A.R. still hold up so well is the way in which it was so closely integrated with the level and audio design, animation work, and the gameplay in general. The enemies feel interactive, reactive, and exciting because so much of what they do relates to their surroundings, as well as to the presence and their knowledge about the player character. The use of calls and responses in their dialogue, and actions like firing while retreating, diving through windows, and hiding make them seem especially dynamic, even if most of those are ultimately achieved through relatively simple audio cues and pre-built map pieces being triggered by the A.I. Also, it may seem like a strictly stylistic gameplay choice, but the ability to slow down the action and see combat unfold more clearly also helps to highlight the quality of the incorporation of the A.I. further.
All in all, while I do think it's fair to argue that more recent games using GOAP or other systems have more advanced A.I., it's the close integration of the enemy A.I. with F.E.A.R.'s various systems that make it so impressive. And with how segmented and often disparate various parts of game development typically are in the modern industry due to larger teams and longer development times, achieving such a tight connection between A.I., gameplay design, level design, and presentation simply isn't going to happen as easily. F.E.A.R. came at just the right time.
46:45 In my view as someone who is currently attempting to build a interesting AI system for a tactical shooter type game I think what's 'haunting the industry' here is how much knowledge is necessary to have rolled up in those few people with real influence of how to allocate the time and budget to see this thru. To know how to craft gameplay like the beloved parts of F.E.A.R you need a high level understanding of how to design the technical state machines, behavior trees, path-finding that is mirrored by the spatial design of the level by someone who is also a proficient gamer who understands how to hold angles and what affordances different cover and placement will provide. The difficulty is in how to create a scenario that forces the player to traverse the environment mindful of exposure and have the AI functionally play 'cat and mouse' attempting to get the drop on one another.
Games of this era was still made by teams of all enthusiasts who built games they themselves would want to play. A small team that can pivot on a dime where everyone understands a lot about the thing they're attempting to build at an enthusiast level are a lot better poised to get it right than a gargantuan team where you have to clear every decision with people who may or may not have the slightest clue what you're talking about or even able to play your own game well.
F.E.A.R itself is very start/stop with those ingenious 'combat space arenas' we all remember being dotted thruout a game intermingled with long segments of the same sort of linear corridor geometry guilty of all the same mistakes how forgettable encounters in titles before/after tend to play out.
Replayability needs unpredictability and that is something even F.E.A.R only succeeds with in a handful of places where the start configuration of the encounter doesn't determine how the rest unfolds. That slice of F.E.A.R that get all of this so right is what needs to be lifted out and studied in great detail, it's a synergy where ones understanding of player tactics, AI tactics and spatial level design all play equal importance.
Thank you soooo much for this, just found out your channel, FEAR was/is an amazing game for me, hearing the mind behind the AI it's so cool! : )
What an excellent revisit/retro celebration, hard to believe it's 20 years ago already! Started with Shogo and although early 2000s Monolith may be a niche it definitely was mine - would be wonderful seeing NOLF3 someday with all the advancements and new discoveries Jeff make on dialogue/etc, including what comes out of the recent language model chatGPT hype. I read his paper back in uni days, the time he spent in the office back then have definitely paid off. FEAR is the mind thriller!
I would love a F.E.A.R Remake in similar style to the Dead Space Remake.
Agreed. Can maybe play Trepang2 in the meantime?
@@AIandGames Hey, thanks for the recommendation! I really love FEAR (and FEAR 2 is also OK) and I played it and its expansions just last year. I'm currently playing TimeShift, it's obviously not as good as FEAR, but it has some neat slow-mo and time warping tricks in it, it's overall pretty good. But I'm definitely giving Trepang2 a try 🙂
@@dodolurker Yeah FEAR 2 is okay, didn't really enjoy FEAR 3 personally (different dev team too). I've never played TimeShift actually. Just noticed it's like £5 on Steam, so might give that a shot. But yeah Trepang2 is 'what if FEAR in 2023'. It's pretty neat.
@@AIandGames TimeShift is... I think one reviewer summed it up pretty well - it looks like a FEAR expansion pack and really wants to be Half-Life 2 😀. Whereas those games are pretty much timeless, TimeShift feels like a shooter from the era. It's also pretty much forgotten today. The only reason I started to play it is because I bought it when it was originally released, caught a glimpse of the box on my shelf and thought "hey, why not give it a go, I didn't finish it back then". But yeah, it's definitely not bad so far, if you can get it on sale, check it out.
I also understand that the GOG version of TimeShift is more patched than the Steam version. It's currently on sale on GOG as well, €9 right now. It also needs another patch, otherwise it crashes when you try to run it, I think it's something to do with too many cores on today's CPUs. But the patched exe is easily found online 🙂
I hope someday someone make it. Would buy it for crazy money
6:01 Ah man, this brings me back to my memory lanes. I was too young to combat the jank and couldn't pass much further than the start, but man I was trying enthusiastically. Good times.
Great video. I've always has great respect for Jeff Orkin and have tried to implement his GOAP stuff multiple times (out of interest mostly, I'm not a game dev, just love the stuff) so it's really great to see him and hear him talk about this. Loved it.
Could see yourself climbing a ladder or running around.... or flying scissor kicking people people in the head in slow mo. How was there never another game like this one. Pure joy.
Small correction. Trespasser (1998) was the first FPS that let you see your character model. Notably health was a heart tattoo on their mammary gland.
I still have my box of Trespasser. It was jank but it literally had heart
I have a video to make about Trespasser that one day we might actually make happen.
In Montezuma's Return you could also see your body when looking down. Not sure if it was the first, but it predated Trespasser for a bit. A much more niche -and weird- game though (and the English version was released later).
Was just playing Extraction Point when this dropped
You and I can hang out.
Jeff Orkin is a LEGEND! I'm happy to learn about him and his experience on creating F.E.A.R. This game will be forever a masterpiece in history.
This is the content I'm here for! 💪
Monolith's 'Blood' had an ad with a guy in a bathtub full of blood, holding two pistols I think, with the caption "Blood. You're soaking in it."
A tongue in cheek take on the Palmolive soap ads of the time.
Masterpiece, kept me involved for months. Still enjoy a run around.
this is one of the best videos on FEAR I have ever seen, if not the best.
I remember a lot of FEAR being held up as a high watermark for AI in video games, still is to this day like in this video, and it was one.
But I remember a somewhat hilarious moment playing it when I would kill an enemy, close a door, another enemy would open the door, I'd kill them, and I quickly had a mound of enemies piled up right at the door that the other soldiers kept running into and and instead of reacting in any sensible way they just kept coming though and dying lol. A moment of artificial dumbness.
Also made me paranoid of climbing game ladders from the jump scares
I'd still love to see a new modern FEAR game.
I think this is the best way to celebrate the 10 years, and I love all the context that Dr Orkin gives us. He really walks you through how GOAP came to be, and I loved that. Amazing, simply amazing video as always!
Alright, time to get my old computer to play FEAR again.
They really need to remake or remaster FEAR like they did with Dead Space.
Sorry to go a bit off-topic but Alien vs Predator 2 was such a good game.
I dunno but it feels like it didn’t deserve the acclaim it should halve.
Both the amazing single player story and the asymmetrical multi player where great games.
Got some great memories of the latter with death matches.
AvP2 was an absolute banger. After getting it up and running to capture footage for the episode, I went back to keep playing it.
I think the first game I remember showing the player-character's full body in first-person was Operation Flashpoint (ArmA: Cold War Assault). I think the ArmA series would make a very interesting video for this channel - the AI had to be able to do their job wherever they were on these huge maps for the time - react to threats of different types, potentially coming from any direction, with multiple units maintaining cohesion
Wow Jeff is awesome. This is a great insight into those early days and how much more exciting and productive the whole industry was.
Having now met the man in person, I can confirm he's a very cool dude.
F.E.A.R.'s AI is still unrivaled in the artform. The Replicas and Armacham forces are so much fun to fight against because the AI was designed to be as smart as possible, thus turning who would normally be generic mooks into dangerous threats whom you have to outsmart.
And one of the reasons why I'm in the camp that also loves F.E.A.R. 2 is that the GOAP AI made a triumphant return, and is still fun to fight against in that game.
“Okay, whatever, it better work” said every boss ever…
That was a really great watch, thanks 🙂. FEAR is definitely still awesome today. It would be very cool if you could track down Tom Leonard, formerly of Looking Glass Studios... I'm still awestruck about his work on Thief: The Dark Project 🙂
I'd love to do something on Thief at some point. I should go and drop Tom a line, he's worked on all sorts of interesting stuff over the years (Portal, Half-Life 2, Left 4 Dead).
My man rocking the 2-Xl on the right shelf. Awesome.
Always love hearing about f.e.a.r
Amazing work once again! ❤
FEAR was one game I had to upgrade my GPU too a 6800 GT. Brilliant game. Absolutely fantastic. Up there with Half Life, Halo and Unreal.
Great video - thank you!
Something I learned from this video is that today’s gaming is missing risk takers. The fact they was using all their resources just to make the AI challenging is truly amazing awhile everyone doubted monolith decisions they excelled by making a timeless masterpiece
Its cool for someone like me to hear about all the work that went into this game when all I can remember is how scared I got almost 20 years ago when you go down that ladder and you see that girl where you just were
Fascinating video! I've always wanted to get into game dev, and similar to Jeff, I was interested in animation and graphics part but AI and behaviour programming sounds so much more interesting to me.
excellent video and interview. you mention around 21:15 that quake 3 was the first to solve or popularize or standardize navigation meshes but I am curious to hear about original unreal's method of solving this problem or if they also used navigation mesh at that time. unreal's bot support was the king for quite a while and even compared to quake 3 has its own qualities. it's possible these bot paths were all hand crafted but I feel like I remember playing custom maps with them. I know nothing and any information regarding the topic would be very cool to learn about
So Quake 3's AAS system isn't the first solution for navigation, but it's largely the one that would later be standardised into other games.
I haven't covered the original Unreal (yet), but I did cover the the navigation system used in Goldeneye 007 a couple years ago which has series of (manually placed) nodes in the world that characters could move between.
@@AIandGames thank you so much for the follow up. appreciate the info
Amazing video. You guys inspire me a lot!
amazing, congrats
51:26 Hey you! You're finally awake. I have no idea why I laughed so much.
Well, thanks to Jeff for making an AI that is fun to setup in combat arenas with the SDK, altho having to manually create the navmesh is something x)
Also amazing video!!
Bring on F.E.A.R 4 pls
Tommy got that original 2XL in the background!
Ok I really want to know how they build the AI.
Time for tea and biscuits for this.
This is why AI today's isn't scary. Not many other games have even come close to matching this AI system.
The fact that Monolith dipped their toes in fps games as varied in tone as FEAR and NOLF....is amazing.
Such wild swings are practically unheard of in today's times.
Companies today are shit scared to dip their toes into anything remotely "wild" and experimental....least they rock the stockholders boat.
AAA gaming today is utterly bland and soul less.....and it doesn't look like it's going to change anytime soon.
At least we can go back to these wild games that scratch that itch.
It's a shame that no one makes games with such high-quality AI anymore. Well Fear, no one has made a game like this to this day.
A.I. drives engagement in games, While graphics have been steadily improving A.I. seems to change in leaps. Some studios just find out what is already out there and is the bare minimum to get the game "working". I hope the future holds some great experiences.
As a hobby game programmer when I was younger without knowing what it even was I made a rudimentary Finite State machine to control my CPU enemy.
I mean you basically listed all the videos i could want from this channel in the first segment: Stalker, Republic Commando, etc
They're on the to-do list. Sadly it's a very long list.
@@AIandGames i can imagine ;) if even just 10% of your videos were this quality i imagine that list will take years.
20:45 funnily enough in my goap implementation i called it GOAT (goal oriented action thinker i explain it) 🐐
I assume so you can then declare your AI is the GOAT?
M a s t e r p i e c e
Love F.E.A.R., love the video
FEAR is known for its amazing AI, and still being one of the most challenging shooters of all time, but. I remember it for the melee combat system. Especially in multiplayer deathmatch, it was absolutely disgusting. You ran extra fast, and your flying and sliding kicks made it so that you were harder to get shot and they one-hit killed.
There's a keybinding to holster your weapon that the game never tells you about. So you can throw down any time in the campaign as well.
You can change the runspeed when you create an MP game, but yea most servers have it really high. But I liked it, the slomo TDM is the most satisfying thing that I could play anyday
Great video as always! SWAT 4 uses G.O.A.P as well, may be worth a look for you too.
SWAT4 is lying around somewhere in the backlog.
@@AIandGames oh sweet! The SDK should be easy to get a hold of and the code is written mainly in good ol UScript. Thank you for all that you share on here, love your work, mate!
I'd love to see what would happen if the AI from games like FEAR, half life etc were body swapped into other titles to see how they stand up
The video is great, awesome insight and i love how he makes it seems simpler than it really is.
Just one observation the sound in the background while Jeff is talking makes quite hard to concentrate on the subject... had to re-watch the video because the background noise is too distracting
graphics are getting and better every year, but not the gameplay... almost everything is so mainstream now
True. Graphics is nice and all but the core of a game is gameplay and that has been lost in current age.
Great video but I had to skip the chapter “Orkin’s path into games’ due to the irritating background noise throughout. EDIT: Seems like it persists in some form for the entire video. What a shame.
I will never understand why they haven't made a Remaster/Remake of the first FEAR game. I am pretty sure they could sell a lot of copies.
this video made me sub.
Awesome video, man! This has to be one of the greatest retrospectives I've seen on FEAR!
Now, I would like to give you some feedback regarding the Brazilian Portuguese translation of the video's subtitles and description. I don't know if you got someone to do it or if you used AI/Machine Translation, but the amount of mistakes and overall lack of care are really concerning. I'm not sure about the quality of the other translations, but the PT-BR one is just not acceptable. In regards to the subtitles, they are too literal and badly formatted for a good reading experience.
I never played F.E.A.R. before, i know i watched videos the AI is very smart, so thanks to this video i bought the game and holy crap, community weren't liying with the AI of this game, a freaking enemy run around a building just to take me by surprise and attack me.
I would like to hear what he would do different now, if he had all the modern benefits of developing a game, that he mentioned and was making a game like FEAR again.
While we cover this in the follow-up episode about Jeff's new venture, the answer is, well... planning, actually...
Unless they patched it, i remember that the loading of the soldiers in an area would slow down the game a bit and all you had to do to clear an area was walk through a door to trigger the enemies and then they would all funnel through the door in a line.
This just makes me wish that NOLF 3 was a thing
Two great men ❤️🫶
My computer couldn't play fear when it came out.
Still haven't but I really need to.
Hell yeah
The incessant droning sound in the background of the second section made it hard to watch.
I would be very interested in your thoughts about the fact that most of the games we recognize as having impactful AI were released so long ago? AI breathes life into games, but these days people seem more interested in using AI to generate silly pictures… Will it come back around?
AI is still improving, its just that the advancements are iterative and less impressive.
these are where the technology that have been improved on started.
With the exception of racing games which use machine learning algorithms for there AI.
This was a great interview! It is sad seeing a lot of AI in games being just kinda of bland and good enough. Disappointing to hear he's doing genAI now. Hopefully it's using ethical training data.
We'll be discussing some of the stuff Jeff's doing now in a future video, but I've seen the tech demo he's built and it's really cool. It's a rather smart application of generative AI, in that it only uses it as part of larger system that still relies on existing approaches (including planning).
And critically, the system relies on users providing their own training data, rather than stealing it from the internet.
f.e.a.r. the fist attempt for an advanced npc ai and also the last..
Comment boost!
Man, I don't have a PC to replay FEAR, Xbox needs to get this back on the marketplace.
If you can find a physical copy it's backwards compatible on Xbox One and Series S/X.
For context: FEAR 2 and 3 were published by WB after they bought up Monolith in 2004. But the original was published by Sierra, and it'll be some rights/royalties issue I suspect between the two that prevents the digital listing on console.
2025 The game begins with a call to the senator. A female voice reports problems with the Source: “Riot. Fettel has taken over control of the prototypes." Armacham Technology was a weapons developer with a military program: an army of clone troopers subordinate to a commander who controlled them using telepathy. Paxton Fettel uses his telepathic abilities to take control of a battalion of clone troopers. Fortunately, to stop an armed conflict, it is enough to eliminate the commander, after which the clones will cease to act and will be absolutely safe. And finding Fettel won’t be difficult: he has a sensor implanted in his head that gives away his location. He will not have any protection, one of the conditions of the program is the safety of the commander, he does not fight along with the rest of the soldiers, but only controls them at a considerable distance. The protagonist of the game is a new operative officer “F. E.A.R.”
Not getting any feedback from anyone ever is a very familiar thing I have experienced myself, working in the industry.
Damn it. We would have had Nolf 3. My favourite Monolith series. This sucks. I don't know why companies look at other companies and decide this won't work even though lot of times a good unique game succeeds. Nolf 2 actually have really good AI. There is an article from Monolith where GOAP is explained with Nolf 2 as base. It is quite clear that Nolf 2 is the start of Fear AI.
Straight jorkin it
🔥
Dude, i am surprised you still didn't make any video on rain world
The backlog is long. Rain World is very far down...
What will happen before 2025 on F.E.A.R.?
Fear remastered and use the money earned to reboot series^^^
None of the developers who worked on fear are there anymore.
Do you trust a bunch of new people to have the same passion for the project, or would they just see it as a product to generate money
back when they made games that made you addicted to palying them
Sorry, it looks good and I'd love to hear the interview, but I can't be in the same room with all the action going on around the borders of the interview video screen. Vertigo and motion-sickness inducing, turned up to 11. Obviously you can't redo this video, but maybe keep it in mind for future ones.
Noted! Thanks for the feedback (and sorry, not something that I thought of when editing it together).
There is the *written* version over on the AI and Games website if you like, y'know... words
As a single-player experience, I'd say FEAR is just about the FPS made since... 2005?
It always makes me sad to listen to people like Jeff Orkin speak about game dev in the 1990s and early 2000s. These guys were always trying to experiment and make cool stuff. Of course they had to worry about sales, but it wasn't like today where every single facet of most games gets degraded in order that the dev can monetize the "enhancements" and "fixes". It just doesn't feel like devs have a passion for what they make, nor any concern for the consumer's experience.
Current game design is not what it was like back in the 90's and 00's. It is getting worse with each passing year.
Mentioning quake 3 without unreal tournament should be illegal
Hey, we love UT around here too. But yes I broke the law.
@@AIandGames Love the videos, live the long form interviews, perfect to watch while working, gonna do my PhD in AI because of you
not nice background sound
Can we get games with good AI again?! F.E.A.R. should have become the benchmark for AI but games seem like they've gotten worse in the last 10 years.
"what if i can make it smart"
I’ll just boost engagement, I don’t have anything to say really that hasn’t been said