The difference between “generative AI” and “procedural generation” cannot be meaningfully nailed down.

Pocketpair Publishing boss John Buckley

Any relation to loss guy?

You’d think that that’s the one thing LLMs should be good at – have characters respond to arbitrary input in-character according to the game state. Unfortunately, restricting output to match the game state is mathematically impossible with LLMs; hallucinations are inevitable and can cause characters to randomly start lying or talking about things thy can’t know about. Plus, LLMs are very heavy on resources.

There are non-generative AI techniques that could be interesting for games, of course; especially ones that can afford to run at a slower pace like seconds or tens of seconds. For example, something that makes characters dynamically adapt their medium-term action plan to the situation every once in a while could work well. But I don’t think we’re going to see useful AI-driven dialogue anytime soon.

You seem to imply we can only use the raw output of the LLm but that’s not true. We can add some deterministic safeguards afterwards to reduce hallucinations and increase relevancy. For example if you use an LLM to generate SQL, you can verify that the answer respects the data schemas and the relationship graph. That’s a pretty hot subject right now, I don’t see why it couldn’t be done for video game dialogues.
Indeed, I also agree that the consumption of resources it requires may not be worth the output.

Isn’t that Garfield’s owner

Pocketpair Publishing boss John Buckley

Any relation to loss guy?

“We might deal in derivative IP, but it’s our derivative IP!”

To be fair Nintendo was heavily inspired by other artists work when designing Pokemon.

Nintendo wasn’t “inspired” by shit. They made an ice cream cone a Pokémon. Keys on a ring? Pokémon. 8 varieties of elemental flavored dog? Check. Oh hey cool look a 2d image on a computer oh wait it’s actually a Pokémon. Dog? Cat? Snake? Bird? Horse? All Pokémon. IMO nothing in Pokémon is actually “inspired”, only ripped off.

Nintendo wasn’t “inspired” by shit. They made an ice cream cone a Pokémon. Keys on a ring? Pokémon. 8 varieties of elemental flavored dog? Check. Oh hey cool look a 2d image on a computer oh wait it’s actually a Pokémon. Dog? Cat? Snake? Bird? Horse? All Pokémon. IMO nothing in Pokémon is actually “inspired”, only ripped off.

But… The developers of Palworld made a game featuring AI generated images.

If you could define a formal schema for what appropriate dialogue options would be you could just pick from it randomly, no need for the AI

It’s been really nice to be able to type a plain question (in any language) into Google and receive a concise answer before scrolling down to confirm with more trustworthy sources. In particular it’s been very good for solving annoyances with UI options by directing me to exactly what I’m looking for. A traditional search will often conflate my search with synonyms (even when using quotations, which is some bullshit), and even ignore what language my search was in.

e: Also you should be careful when clicking on any links provided by an LLM because they can accidentally send you phishing links.

As an amateur game dev, I believe AI will crash out for the public before it becomes truly useful for programming. I’ve heard colleagues try to use AI , but it often just creates more work. When the AI doesn’t know the answer, which is often. it makes something up, leading to errors, crashes, or hidden issues like memory leaks. I’d rather write the code correctly from the start and understand how it works, than spend hours hunting down problems in AI-generated code, only to never find the issue. Full disclosure I use Chatgpt to edit my dialogue as my English is not great.

The difference between “generative AI” and “procedural generation” cannot be meaningfully nailed down.

Companies can change their mind about stuff like this.

Yes. Except they’re still selling that game.

It would not be a fully determining schema that could apply to random outputs, I would guess this is impossible for natural language, and if it is possible, then it may as well be used for procedural generation. It would be just enough to make an LLM output be good enough. It doesn’t need to be perfect because human output is not perfect either.