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.
FEAR is a high watermark for AI in video games, but it's also a high watermark for
marketing AI in video games.
They sold everything that the game characters could do in pitching the game to the audience, and they used techniques of audio and animation and lighting to ensure that players did not miss the AI doing its tricks. (It's also a high watermark for putting the extra work into programming unique pathing incident routines and level design specifically crafted to maximize enemy character choices in encounter spaces to give a great playspace for the ideas pioneered in the AI design.) "Combat Adaptable A.I." was a bulletpoint on the box, and was a key talking point in every preview. Games are rarely funded these days with such singular focus on one aspect like this.
As Orkin says in the interview, it's not that the techniques of FEAR were only used in FEAR, or were never surpassed over time. (In any case, the game use of GOAP is inherently hard to "surpass" since it's generally human design and academic research which powers the FEAR AI rather than generational leaps in hardware horsepower. It helped that they had the machines they did then just as they made this breakthrough, but more CPU power wouldn't completely mean more intelligence for characters, it would mean more work on the designers' parts to come up with additional intelligent things for them to do.) You see FEAR's progeny in gaming all the time. What you rarely if ever see, however, is game design and production focus placed so specifically on highlighting and exploring the capabilities of this (or more robust) artificial intelligence approach. It was a golden era for taking a good idea all the way to the end of the line, and FEAR had those kinds of good ideas and people with the time and experience to know how to use them. Games have technically gotten smarter since then, but gamers and game producers have not valued "smart" enough since then to make it a key reason to get hyped for new games.
For those of you who don't know, F.E.A.R is an acronym for First Encounter Assault Recon and a great example of how AI can be used in video games.
modl.ai