Hello everyone,
It is perceived that gaming industry is more inclined towards multiplayer games like Fornite. Except Sony not many publishers appreciate immersive single player experiences. It seems for me, That market value of gaming AI developers is lower in comparison with other areas of gaming career like computer graphics.

(1) Do you agree with me that gaming industry lacks interest of innovating AI, and that market value of AI developers is lower? (2) Why do you hold that position (whether agreed or not) (3) Where do you think the right direction the industry should focus on?. Feel free to add any additional commentary.

The industry, as every other industry, is interested in first place to maximize the financial value of their games. I'm not at all talking about indies but the AAA companies. As in every other industry, people are doing work for an acceptable amount of payment and those sallaries have to be generate in a capitalistic system.

Game AI is expensive as you need experts that are capable of creating it where multiplayer is more popular these days and eSports is giving another platform of extra financial value while it is not necessary to invest into automated gameplay, like a rich AI instead of focussing on the PvP experience.

So what it is used for today is simply to control some NPCs to interact or mostly kill without any strong template ohter than

• spot
• attack
• defeat

and that's it. They're not even putting much effort into different difficulty levels. Unreal Tournament for example simply increased the NPC damage instead of making the AI clever at different levels

mostafatouny said:
0) That market value of gaming AI developers is lower in comparison with other areas of gaming career like computer graphics.

1) Do you agree with me that gaming industry lacks interest of innovating AI, and that market value of AI developers is lower?

2) Why do you hold that position (whether agreed or not)

3) Where do you think the right direction the industry should focus on?

0) That market value of gaming AI developers is lower in comparison with other areas of gaming career like computer graphics.

You've got to be really careful about terms like “AI”. The term means radically different things to different people.

AI in games is quite different from AI in research and advanced tech, because games are not about machine learning and solving complex problems. AI in games is to create a fun challenge which the player can overcome. Unlike machine learning situations where complex self-organizing systems are used, game AI needs to be easily understood by humans tuning it so they can adjust difficulty.

Yes, game AI is not a cutting edge field like computer graphics. Computer graphics will always require some individuals who are current on this month's latest and greatest techniques, where AI can continue to leverage decades-old methods and techniques. Therefore, game AI programmers are generally in lower demand than cutting-edge graphics programmers and have less market value.

1) Do you agree with me that gaming industry lacks interest of innovating AI, and that market value of AI developers is lower?

Yes, absolutely, and for the reasons stated above. If we used deep machine learning techniques the only people who could win are other AI's and bots. It is extremely easy to write an AI that is adaptive and can utterly destroy human players. Those AI's are not fun.

Our industry is looking for AI that is easily tuned and adjusted, where designers (not mathematicians or data scientists) can make tiny adjustments to create difficulty levels that ramp up from very easy to insanely difficult.

2) Why do you hold that position (whether agreed or not)

I've worked on enough games to see it happen in action. The first few rounds of AI tuning range anywhere from trivially easy to quite literally impossible. In both cases the values need to be adjusted to find the fun.

3) Where do you think the right direction the industry should focus on?

The industry should continue to maximize fun and entertainment.

@mostafatouny cuts development time / debugging to leave out AI. Also reduces performance costs letting them reaching a broader audience. If you are developing for more capable hardware, there's no reason to not have great AI unless there's no budget for it.

frob said:
Yes, absolutely, and for the reasons stated above. If we used deep machine learning techniques the only people who could win are other AI's and bots. It is extremely easy to write an AI that is adaptive and can utterly destroy human players. Those AI's are not fun.

This bugged me for years, even with the basic AI's we have now. Generally as the difficulty is turned up I generally find them unfun due to winning in ways real players simply can't.

In many FPS higher difficulty just sees the bots still walk around like idiots, but at some point they just instant hit you even with both “players” moving erratically. And just generally playing against them on any difficulty is nothing like Player vs Player, right from the bottom rank up.

In strategy games they often resource cheat like mad, ignore fog of war, and can micro control every unit even on different parts of the map (in an RTS), no being surprised, etc. And the AI generally pretty much lacks any adaptive strategy to make use of the terrain, exploit player mistakes etc.

If some sort of deep learning or other technology let an AI compete on a more level playing field I think would be a lot more fun in general rather than just either cannon-fodder for the single player experience to rack up 100's, 1000's of kills, or for some generally not very fun multiplayer AI.

So no resource or stat bonus, or other cheating. Only seeing and controlling what is on a virtual “screen” + minimap and other-UI, with a human like reaction-time and precision to move about and take actions. Of course still needs to be a difficulty level, an “ai ranking” of sorts.

EDIT:

Also maybe just better AI for player-controlled stuff as well. It can be infuriating when stuff decides to just act in a dumb way, and get it's trapped or killed.

(1) Do you agree with me that gaming industry lacks interest of innovating AI, and that market value of AI developers is lower?

The attitude certainly doesn't surprise me. Like other people in this thread have mentioned, people (investers and executives, but programmers and artists as well) are responsive to where the money is, and that will always shape the game dev industry.

(2) Why do you hold that position (whether agreed or not)

I've spent enough time watching the sausage get made in various software industries.

(3) Where do you think the right direction the industry should focus on?

I guess I mostly think that the game development industry might have to wait for AI to be pushed forward in another industry. There are other industries where it makes sense, maybe in terms of online shopping services or military software or something, and the work in those places will eventually spill over.

Another possibility is if some smaller game company creates a game that spawns a new genre, one in which cool AI has a huge impact on how fun the game is. I have no idea what that would look like. Maybe something where the enemy creatures or landscape evolve in completely unexpected, generative ways? Or where there are certain areas that you need to be powerful enough or skilled enough to enter, because you're only going to have one shot at the clever AI-enemies?

The only other thing I can think of is if the AI could be designed in such a way as to really guarantee that a game has endless replayability. I think that replayability is actually one of the things that makes games like Fortnite so profitable: Same graphics, same limited maps or map variations, same game software, but people keep coming back because it's "different" every time. If you could create a similar replayability dynamic but with AI (or a mix of AI and PvP), then the money might start to flow in that direction.

Totally agreed that AI in research is different from one in entertainment. The former aims for rational problem solving, while the latter aims for a psychological state of fun. Check this video

Nonetheless, A very recent research emerged for entertainment-centered AI. AIIDE is journal specifically for that. Here is a paper from IEEE.

Even if gaming AI relies on ad-hocs unlike AI in standard research, R&D departments in gaming industries could have invested in those ad-hocs yielding innovative gaming experiences. indie game RimWorld innovated AI storyteller I had never seen before. What could have been done by AAA budget games then?

There's a few things of note here… first off, the premise of the gaming industry lacking enthusiasm for game AI because of the focus on multi-player games is potentially (and historically was largely) upside down. For a good while, people were drawn to PvP because AI sucked and people wanted a greater challenge. Companies really tried to crank up their AI once we got to the “OMG it sure is pretty to look at but really dumb” stage of NPCs.

So why, then, are we still headed towards online PvP games so much? Well, there are a variety of reasons – the social aspect being one, for example. However, there was another premise scattered throughout this thread that wasn't completely broken, but on a parallel track. People alluded to how expensive it is to create AI and that set studios back monetarily. While yes, we AI programmers and the designers that use it don't work for free, that's not the problem at hand. The monetary outlays for hand-authored, usually single-player content are not in the AI… but rather in the content creation itself. If you have a lot of different types of NPCs, they not only have their own artwork and animations, but really have a lot of their own storylines that have to be written, designed, and especially VO recorded for them. THAT is expensive and takes massive teams of people to do. Sure, the AI has to support that sort of stuff and the designers need to leverage the AI – so there is some interaction there. However, if you look the team size, number of hours (days… weeks… months… years…) that go into content creation team for a single-player game compared to the corresponding numbers for the AI team, it is a huge gulf.

The attractiveness of online games at that point is that the players are providing the plot. In fact, much of the experience as a whole. Yeah, there are levels that are laid out and characters that are rigged, animated, and decorated… but, on the other hand, you can have 500 people in a game that are all using minor variations of the same bloody character. You don't have to have a team of people spending huge amount of time saying, “and orc #214 stands here, moves here, says this (which is usually a monster dialog tree), and drops this special loot for quest #32b.”

So no, there isn't a “lack of interest” in AI in the game industry. As a co-founder of the AI Game Programmers Guild and the co-founder and host for the first 10 years of the GDC AI Summit, I'm very familiar with hundreds of AI devs around the world and what they have been working on and what they are working on still. Some companies are more interested than others – but that is largely because of the type of games they put out for a different type of genre and/or audience.

mostafatouny said:
Nonetheless, A very recent research emerged for entertainment-centered AI. AIIDE is journal specifically for that.

This confused me… are you suggesting that AIIDE is “recent”? And that AIIDE is “a journal”?

Some of the most interesting AI research, period, has happened within the context of gaming.

Deep Blue winning over Kasparov; Watson winning Jeopardy; AlphaGo winning over Lee Sedol; and AlphaStar winning over Team Liquid and placing in Grandmaster level in StarCraft II.

I think the question is more what people want in games. Do you actually want a fake opponent? If so, do you want that fake opponent to be really hard to out-smart? Many modern games are much more like interactive movie experiences, than a traditional “board game” of two opposing sides battling it out. “Call of Duty” single-player is not about two sides in a war battling with well-defined rules and limitations; it's about running the player through a number of scripted scenes that deliver a particular experience. This is fine; it's very enjoyable to many people! And once you've played through it and eaten content that took three years to produce in the span of fifteen hours, there's another game to play with your free time, so re-playability isn't really a problem.

Then there's multi-player. Game developers have the challenge of a bunch of players that want to experience ever-changing content. What better way to solve the problem for all players at once, than by pitting them against each other?

For “modern AI” to be able to perform at the level of “modern scripted scenes,” we'd basically need an AI that's so good that it could write a coherent novel. That kind of AI doesn't yet exist, even in the case of leading edge systems like AlphaStar. People's expectations are simply too high; “versus” games with a good opponent is a tiny fraction of the market, so a “better versus AI” simply doesn't sell many more games, compared to “ten more hours of scripted scenes.”

Some people, especially those who prefer “versus” type gameplay, and who don't like having to wade through the sewage and trolls found in online games, may think that this is the wrong focus, but the market overall (e g, “all human beings who pay for games”) disagree.

@iadavemark,
Well, well, .. One of the industry leaders replied to my post. The pleasure is mine Mr. Dave Mark. Let me seize this opportunity to congratulate you on your book Behavioral Mathematics. I liked how you balanced accessible reading with scientific rigor. It Certainly fulfilled a gap in the literature.

This confused me… are you suggesting that AIIDE is “recent”? And that AIIDE is “a journal”?

Thank you for your correction. It's my bad. I meant by recent entertainmentand-centered AI the recent growing of Affective Computing field. It aims to let AI interact emotionally instead of rationally. It emphasizes how machines are adapted and engineered towards humans psychological well-being. Most famously, It is applied on health-care and Autism. However, Some papers are aimed on virtual agent and even mentioning video-games. Oxford Handbook of Affective Computing, The most modern monograph for researchers devoted a whole chapter named Emotion in Games. Here is a snippet of the paper's abstract:

One of the most promising and also challenging applications of affective computing research is within computer games. This chapter focuses on the study of emotion in the computer games domain, reviews seminal work at the crossroads of game technology, game design, and affective computing, and details the key phases for efficient affect-based interaction in games.

Second. The correct recent journal is IEEE Transactions on Games not AIIDE conference.

If I got your point right, Then the great interest in PvP is due to the poor state-of-the-art AI NPCs and the high hand-authored content-creation costs. But it is not the case that the low enthusiasm in gaming AI lead to the rise of multiplayer games. You ensure that AI in games still have its own interest within the community, But the obstacle is in generating business profit.

Let me seize the opportunity of having your valuable opinion. If publishers find multiplayer games more profitable and find single-player games with hand-authored content expensive, Does that indicate the low funding in gaming AI? I think a scientific research gets funded due to its potential in having impact on industry. I remember the recent debate of Whether CERN should build a new particle collider, As it might not have an impact on engineering/industry. So, if gaming AI is not promising for publishers in making profit alike multiplayer games, I see no motivation for them or even universities to invest on it. Let me know your opinion about that.