We're working with a relatively large codebase, much of which could be called legacy and much of that legacy we're still building on because we still have a lot of other tech debt that's prioritized higher for fixing but still need to further develop the system, and AI is simply unreliable. It's not necessarily bad, but it can't understand the codebase the way humans can, so it makes dumb mistakes because of that, but from time to time it also makes dumb mistakes unrelated to that. I'm sure it would do better in a smaller project or a project that was built to higher standards from the ground up, but unfortunately it's not.

For what we're doing, the risk of any bugs resulting from poor AI code are simply such that basically everything the AI does needs to be reviewed, preferably carefully, lest we risk stupid mistakes getting through. It's slow work because of that, maybe still somewhat faster than doing it manually but reviewing so much code is not fun.

We're using paid versions of several models, I believe, and that includes Claude. I think the models available to us are top-notch, for what is available at the moment at least.

I'm glad it's working out for you, but for us, there's a long way to go still before AI can really be a game-changer. For us specifically, I feel like it's more like a minor convenience at best at the moment.

Yeah, AI NPCs is one thing I'm looking forward to, if they're done properly - after all (and this is something that most people either forget or don't know), a lot of early RPGs from 80s (at least made in west) had text parsers that you could use to ask NPCs about all sorts of thing you can think of. Of course, those NPCs were dumb and scripted, so interaction with them was very limited.

"Holly Grail" of NPCs that you can talk to is not new and comes directly from TTRPGs were DM/GM represents world and NPCs in it. AI is ultimately tool to try to replicate that experience in video games. Of course, this will (hopefully) change how those games are developed, with VG RPGs becoming more akin to their tabletop cousins, with actual "smart" NPCs, instead of scripted quest dispensers with quest marks hovering over their heads. thus interacting with them becomes actual "social encounter" from TTRPGs, instead of "just give me the quest/info" thing of VG RPGs.

I don't know. If I wanted to have a real conversation I believe I would just talk to an actual person, lol. It's usually not what I'm after when I'm playing video games. So I'm not too excited about it. However, other people have other tastes and that's fine. So if AI leads to some new gaming experiences that people enjoy I'm all for it, even if I wouldn't play it myself.

AI should only be used to do mundane and repetitive tasks

It's not really about conversing with NPCs - ask yourself with how many people you actually converse in real life, versus with how many people you have brief chats to get some info.

Imagine classic "you enter the inn" scenario - in most modern RPGs you have NPCs who are sitting there and who have no other task but to be a backdrop. Then there is one or more that have some info, or quest. Compare that to TTRPG approach where you enter the inn and all NPCs are equal, at least in theory. It's up to you to dig up info you need, or in conversation with some to discover something new you didn't know - some of early CRPGs from 80s actually were like this, but with dumb NPCs, so it was not quite rewarding experience.

Now, I get that some people don't like whole "social interaction" pillar to be nothing more than quest/lore dumps and to get it over as quickly as possible and get on with action - but for folks who like that 3rd pillar to be as important as other 2 (combat and exploration), clever NPCs via AI are quite important addition (if done properly).

DLSS, FSR and PSSR before that DLSS 5 shit were completely fine things.

The problem is not the IA, is how you use it.

Which isn't a criterium against AI. You can use AI and still curate the results and strive for high quality. Obviously - living i capitalism - most companies opt for low quality and cost reduction. With or without AI.

EDIT:

Let me elaborate on this. You can have high quality printing done with care. I recently purchased "The Ultimate Guide to Rebuilding Civilization" (not meant as advertisement, just an example), and this shows how high quality printing is indeed possible. But if we look at history, than for the most part the printing press lead to less high quality carefully made beatiful manuscripts, and lead to a lot of very low quality prints. This isn't a result of the abilities of printing technology. It is a result of societal incentive structure. Don't take this as a plea to return to pre-printing press technology, I just state things as they are.

Similarly you can still put the same amount of care into stuff, while using AI. I am a programmer, and if you use AI you can use it in a way that doesn't increase your code output, or not by much. But with the ability to explore more avenues, try more prototypes, solve more problems you can with proper care produce code in higher quality. But that said, if you want you can vibe code thousands of lines of code a day. Obviously nobody can check and quality control this much output. But you can and people do, and as this output is much more than any human can do I expect that vibe code will overtake human produced code in a matter of months — if it hasn't already happened.

As an example how bad such badly curated code can be:

https://www.youtube.com/watch?v=mJ2GZRV63TE

Stuff like this is funny. But to be real: for a lot of code it doesn't matter if it is bad. Some random website, simple tools for a company and so on. There is no real problem. While a programmer like me can laugh about stuff like Garry's list, but in the grand scheme of things it doesn't matter. And AI allows people that can't program and can't pay a coder to create something they need. Bad yes, but a bad thing is better than no thing often. So AI is empowering for non-coders. And people who care, can still create quality. The incentives of our society again are in the way. If programmers get paid to produce stuff, the incentive are on quick and cheap implementation, and AI improves these goals. So most commercial code will be soon be produced by or with help of AI with lower quality. But again, not a problem of the technology, a problem of society. 

EDIT 2:

About the amount of code produced, there are already quotes from famous programmers, that tackle this and were made, before AI made it possible to produce torrents of code.

“The code you write makes you a programmer. The code you delete makes you a good one. The code you don’t have to write makes you a great one.” – Mario Fusco

“Measuring programming progress by lines of code is like measuring aircraft building progress by weight.” – Bill Gates

“One of my most productive days was throwing away 1000 lines of code.” – Ken Thompson

I said this about a year ago, but I’d be totally fine with AI being used to let us respond to NPCs however we want, and the NPC would dynamically generate a response back to us.

In a lot of RPGs, especially ones where your character is meant to be a self-insert, you’re usually given multiple dialogue options when talking to NPCs, but too many times those choices all feel basically the same. No matter what you pick, the response you get back from the NPC is usually more or less identical.

AI could actually make those interactions feel more dynamic and personal, instead of just giving the illusion of choice.

They could do similar stuff like that before "A.I." chatbots existed, as SvennoJ explained. And it didn't require plagiarism and a massive waste of resources far beyond any other consumer electronics.

"A.I." should at most only be used for scientific/medical research, and that's it. It shouldn't be Silicon Valley's favorite new toy to wreck everything and destroy jobs (like in art & voice performances, among countless others) in a last desperate attempt to make the line go up. I don't particularly care what the "A.I." apologists have to say on the matter.

As of now they are still completely fine, AI upscaling is in general a good way to keep improving graphics, whilst games remain playable.

The problem with DLSS5 is that it fundamentally changes the source material. However if DLSS5 is optional, I don't really see the problem with it. Maybe some people prefer the look, the environments looked better and didn't fundamentally change all that much. In a FPS game without a lot of faces DLSS will probably still be a big improvement.