90% of Games Developers Already Using AI in Workflows, According to New Google Cloud Research

I mean... I believe it.

No, not necessarily the intended way. But in terms of tools.

Like, one of the coolest things in an image editing program is the "magic wand" and Adobe's suite goes above and beyond to genuinely feel like magic where you can outright erase people or procedurally weather stuff and so forth. And... that uses most of the same code paths and "tech" as "AI"

Same with a lot of audio programs where you are spending a lot of money to buy plugins/algorithms to handle a lot of the heuristics when blending different sounds. Let alone using asset packs that likely are pure AI Generated Content.

And same with coding. Microsoft et al increasingly shove "AI" tools down our throats. But even just googling you are going to get gemini bullshit that summarizes the stack overflow page you are about to read. Or you just use chatgpt to remove the middle man entirely.

And considering that the term "AI agent" has increasingly become a buzzword specifically for these purposes (sometimes it is a full LLM. Sometimes it is just an old school neural net. And so forth)? Let's look at that breakdown on slide 7

44%: Asset or content optimization: Pretty much exactly what I mentioned regarding Adobe and the various audio mixing programs
38%: In-game Coaching or Automated Tutorials: These I suspect are "real" usage but are mostly geared towards proof of concepts. Although I could see an argument that "I have 77 skill points that are unallocated" pop-ups back in Guild Wars were "AI" because... they were. They were just incredibly rudimentary checks of if skillPoints > 10: whineLikeALittleB()
38%: Dynamic Balancing and tuning of gameplay: See above but think "The Director" from L4D or just lowering the difficulty the tenth time you die against a boss
37%: Procedural (...) generation: This is procedural generation. That is, and has been, "AI Generated Content" for literally decades
36%: Adaptive difficulty: See In-Game Coaching
35%: Automated testing and bug reporting: Honestly? This should be at 100%. There is absolutely zero reason to not use a (locally run) tool for this kind of stuff. You just use that in addition to the QA team.
34%: Advanced NPC Behavior: I suspect this is mostly the aformentioned In-Game Coaching but also buzz word games like Star Citizen
34%: Internal Studio Functions: Them's be lying considering how much "AI" gets shoved down our throats when you use MS Teams or Google Cloud or whatever.
33%: Real time voice or audio enhancements: Homie, that's auto-tune. Again, this is probably closer to 60 or 70% and those folk just don't realize what the tool they used to "clean up the audio" was doing

But yeah. This is clearly marketed as the idea of "They asked Gemini to make a game for them" rather than "They used the existing, and actually reasonable-ish, tools to do their jobs". Which gets to the idea of "AI" to enhance workflows rather than to become it.

7 comments

The company pushing AI claims 97% of developers use AI. That sounds like a number AI came up with.
- As a full stack web dev I've been using AI off and on and while it seems really good at basic markup language, it almost never gets PHP or JS right.
  It suggests weird ways of doing things that either don't work, are completely asinine, use some random unknown libraries, or is just outdated.
“See? This developer used an AI to help him work out how to resolve this Unity error”
Meanwhile the dev: “Wtf? I just wanted to do a regular search for that error, not AI.”
- You fell victim to one of the classic blunders - the most famous of which is "never get involved in investing with a tech bro" - but only slightly less well-known is this: "Never expect a search not to become an AI prompt online!" Ahahahaha
I mean... I believe it.
No, not necessarily the intended way. But in terms of tools.
Like, one of the coolest things in an image editing program is the "magic wand" and Adobe's suite goes above and beyond to genuinely feel like magic where you can outright erase people or procedurally weather stuff and so forth. And... that uses most of the same code paths and "tech" as "AI"
Same with a lot of audio programs where you are spending a lot of money to buy plugins/algorithms to handle a lot of the heuristics when blending different sounds. Let alone using asset packs that likely are pure AI Generated Content.
And same with coding. Microsoft et al increasingly shove "AI" tools down our throats. But even just googling you are going to get gemini bullshit that summarizes the stack overflow page you are about to read. Or you just use chatgpt to remove the middle man entirely.
And considering that the term "AI agent" has increasingly become a buzzword specifically for these purposes (sometimes it is a full LLM. Sometimes it is just an old school neural net. And so forth)? Let's look at that breakdown on slide 7
44%: Asset or content optimization: Pretty much exactly what I mentioned regarding Adobe and the various audio mixing programs
38%: In-game Coaching or Automated Tutorials: These I suspect are "real" usage but are mostly geared towards proof of concepts. Although I could see an argument that "I have 77 skill points that are unallocated" pop-ups back in Guild Wars were "AI" because... they were. They were just incredibly rudimentary checks of if skillPoints > 10: whineLikeALittleB()
38%: Dynamic Balancing and tuning of gameplay: See above but think "The Director" from L4D or just lowering the difficulty the tenth time you die against a boss
37%: Procedural (...) generation: This is procedural generation. That is, and has been, "AI Generated Content" for literally decades
36%: Adaptive difficulty: See In-Game Coaching
35%: Automated testing and bug reporting: Honestly? This should be at 100%. There is absolutely zero reason to not use a (locally run) tool for this kind of stuff. You just use that in addition to the QA team.
34%: Advanced NPC Behavior: I suspect this is mostly the aformentioned In-Game Coaching but also buzz word games like Star Citizen
34%: Internal Studio Functions: Them's be lying considering how much "AI" gets shoved down our throats when you use MS Teams or Google Cloud or whatever.
33%: Real time voice or audio enhancements: Homie, that's auto-tune. Again, this is probably closer to 60 or 70% and those folk just don't realize what the tool they used to "clean up the audio" was doing
But yeah. This is clearly marketed as the idea of "They asked Gemini to make a game for them" rather than "They used the existing, and actually reasonable-ish, tools to do their jobs". Which gets to the idea of "AI" to enhance workflows rather than to become it.
- Note that game dev is a setting where both users and developers already tolerate a fair bit of jank or bugs, and where having ideas is relatively cheap but iterating on them is not at all. It makes sense as a fit.
Yeah more AI slop games I would never play.