When AI will be able to generate credible stories with credible characters then put them into credible worlds whose data can be fed to a Open Source game engine, we'll hear a great disturbance in the Force, as if thousands of game designers and publishers suddenly cried in terror and were suddenly silenced.
On a serious note, it seems to me technology isn't that far from that point.
> On a serious note, it seems to me technology isn't that far from that point.
It feels at least a couple of decades away.
I think the most compelling use for LLMs and generative AI right now is parsing massive tomes of knowledge, which can then be summarised at use-time.
Like imagine pilots being able to consult an LLM with all of the aircraft's blueprints, technical reports, operational history, etc. during flight. Likewise for spacecraft when debugging in flight.
But the internals really need to go beyond just token prediction, to being able to use Abstract Syntax Trees for programming suggestions with correct types, or building ontology trees for answering queries factually. The former is a major frustration with Copilot (when it gives relevant suggestions at all).
chatgpt was released 1 year ago, the transformer is 7 years old, I feel quite the opposite, and also I am quite excited and frightened at the same time.
Have you seen Sora lately? Every few months we get a new big AI innovation or major improvement. We are not in an end of Moore's Law phase in AI, we essentially just started, and the growth is currently exponential not linear.
> On a serious note, it seems to me technology isn't that far from that point.
Perhaps, but because it is hit-driven, I can easily believe "thousands" being impacted even when the top-10 appears unchanged by the reaching of this standard.
Credible != compelling. AI generated stuff is garbage, it's so far away from generating something exciting, it's as far as it was 10 years ago, the only difference is that now it is able to generate garbage (instead of doing nothing).
Even a kid can have better ideas than an AI fed with billions $ of nVidia hardware and data coming from unpaid labor.
It's really hard to imagine that an AI-generated storyline would be worse than say most Bethesda questlines of the last 10 years or so. Or say Ubisoft circa AC:Valhalla. There are so many storylines and quest objectives in games these days that are more or less:
Go here and get a pencil. <watch a cutscene>
Ok great. Now I need 3 more pencils. <watch another cutscene>
OK now go here and follow an NPC for ages while they walk really slowly. <watch another cutscene>
OK now shoot the guy <watch a cutscene>
OK we're done, have a cookie. Any time you want to, come back here and I'll send you to go fetch more pencils.
I only had the blue pencil after the first quest. I had to go back. Not for the cookies really, but rather because I needed the yellow, purple and especially the red pencil. Only then would I be the level 10 mage capable of retrieving the black pencil.
You act like they were making a fool out of me or something. /s
There was a parody video a while back, can't find it now, about RPGs from the NPC's perspective. PC steals all their stuff and also their grandmother, ends up receiving the sword of grandmothers.
I don’t know. There are a lot of popular games that are basically made to a formula. That said, I don’t see this being possible any time soon for non programmers (with the aid of ai). Consider the amount of debugging needed.
If the computer can actually program coherently, without bugs, and understanding the program as a whole and how the pieces fit together, then it can program itself.
I don’t see this being possible any time soon for non programmers
I don't see this being possible at all for non programmers. If AI companies have models that can generate games in 5 years, there's no way they will let other people use them. The companies with game generating models will simply become the developer and publisher.
If they can generate games, especially those deliberately addictive casual games, then they have a money printing machine. They'll just hire psychologists, economists, and testers to see which games are the most addictive and extractive- er- umm- I mean, which games are the most fun and compelling. Then they'll release those games. Further flood the games market with tripe, and probably get pretty rich doing it.
I hear such comparisons quite often on discussions like this, and I always wonder… have you seen a kid's idea of "creative"? Even one about to finish school?
One time my geography homework was a cut-out of a photocopy of an encyclopaedia image of Italy, surrounded by notes copied verbatim from that same page, and it got top marks. My GCSE English "creative writing" exercise was a mixture of the empty village in Riven, Pratchett's Granny Weatherwax, and the "oborot" shape-shifting spell from Sabine Baring-Gould’s Book of Were-Wolves. When we were tasked with coming up with our own versions of the witches' brew from Macbeth — why is Shakespeare sill so popular? Don't we have any new ideas? — all the students came up with much the same "new ingredients" as each other.
I think all of this AI generated content still has a ways to go until it can be stand alone content, but when it is being used as a supplement to human creation, it can be fairly useful to either generate a starting point, or to fill in the blanks.
BUT...I am seeing a lot of AI generated content pop up in graphic design(emails, work decks, etc...) and some of it is terrifyingly bad.
Humans create shit stories too, and at scale! Just look on Steam at the deluge of shit that humans create.
Less is more.
With that said, I agree that it could be awhile before an AI generates something compelling.
Edit: As an aside this touches on a capability that I was wondering about. Suppose you use Sora to generate a video, how do you fix the issues? It would be nice to rather than generate a video, generate an animation project or something similar so the issues can be fixed more easily. Generating video games seems like a step in that direction.
On a serious note, it seems to me technology isn't that far from that point.