If you want to lecture people about how bad it is to be using Image Gen, then there’s another more popular thread for that 
In any case, I likely wouldn’t do it now, but I did it then.
I’m sorry if I came off as lecturing about the evils of image generation. My intent was to make a point about the nature of generative AI and that you can’t just isolate a part of the training set like you were talking about. I use generative AI a bit myself, just not for creating IF.
4 Likes
The thing that scares me is that authors would have this part of the game be out of their control. You’d get messages like “how do I find the gold Bob needs to buy charcoal for his backyard barbecue” when you did not write anything about gold, charcoal, or barbecues, but the AI hallucinated it.
Players who are out of ideas tend to look at all the text more specifically for clues, and LLM descriptions tend to be long and flowery, which can throw them off the path from things the author actually did implement. Talking to an NPC should be a bit of flavor text or for puzzles the author intended, not lead you to hunt down things that aren’t there and distract from the main point of the game.
I don’t even think AI prose with heavy restrictions would be all that useful: at that point, you could just write it yourself to avoid letting any content you don’t want sneak in. People know when dialogue is boring and how to make it interesting: AI doesn’t. Also, with LLMs, people could easily derail the NPCs with controversial topics. I just think the risks for both players and authors are too high for this to really be worth considering.
7 Likes
Yeah, those are all very real pitfalls with the current state of the tech, but even if fixing them requires another paradigm shift in how AIs work algorithmically, a significant advance in classical computing power, the extreme parallelism of quantum computers, or some as of yet undiscovered mode of hypercomputation, the tech is only going to get better as time goes on, and while even the experts can’t give us any good estimates of timelines, it’s probably only a matter of time beforereplacing NPCs with LLM-type chatbots is viable.
1 Like
I have little hope for AI “getting better”; it’s just the next tech fad that’ll blow up and die out. Remember virtual reality? NFTs? The Metaverse? It’s already starting to scrape other AI-generated slop and artists have been glazing to poison the dataset. People overestimate what it’s capable of, and actually trying to put it into practice highlights why it’s not viable in the long term. AI has also already gotten a lot of pushback, and an association with low-effort slop content. It’s not going to be the big thing techbros think it is.
7 Likes
(rubs temples, recalling annoying memories)
Remember when marketing talk-heads were predicting it would replace most forms of gaming and computer interfacing?
I wonder if there’s survey data out there about how long it’s been since VR purchasers last used VR.
4 Likes
I have an Oculus that I’ve played with exactly once, and that was two years ago, and a friend told me theirs has been in its box for years because they use it as a stand for their mirror. Most VR games simply being tech demos that lose their luster after about 10 minutes didn’t help its longevity either.
3 Likes
7 Likes
I used my Occulus to play Moss, and haven’t used it for much since. Mrs Hituro uses it to play Beat Sabre quite a lot, and tries other games from time to time — usually in the same rhythm sort of category.
Then again, the re-released Riven is available for the Occulus, so …
5 Likes
Oh, I wouldn’t be surprised if ChatGPT and its ilk end up collapsing under their own weight and AI fades from public consciousness for a while. But AI as an area of research dates all the way back to Alan Turing and the days when electronic computers were the size of buildings and Asimov’s Three Laws of robotics predate ENIAC. Even if the current paradigm proves a dead end towards creating an AI that is truly Human level at all mental tasks and Silicon Valley loses interest in trying to turn a profit on current generation AI models, AI as an area of active research isn’t likely to go away any time soon.
As for VR, the technical limitations might have prevented it from truly taking off even if the 2010s version completely and utterly obliterated the 90s version, but I fully expect someone to give it another go in another 10-20 years. And I say this as someone who mostly didn’t care since 95% of the tech focused on the visuals and what experiments there were with adding tactility and sound had prohibitively expensive cost of entry.
NFTs were just a plain bad idea and more interesting from a mathematics standpoint than as anything with practical use.
No idea what the Metaverse is.
Oh, and on a somewhat related note, remember PDAs? You might not realize it, but they are another tech idea that was too limited and too expensive in its original form, but which now are so ubiquitous that pretty much everyone and their grandmother expects you to have one, more people have one than have a desktop computer these days, and outnumber people in some parts of the world… Of course, we’ve been calling them smartphones ever since Apple made one with mainstream appeal in the form of the iPhone.
7 Likes
AI as a discipline is going nowhere, I imagine. Since I did my masters in it (a thoroughly terrifying 25 years ago) there has been steady progress, and AI techniques are highly integrated into modern technology. From predictive text on phones, handwriting recognition on tablets, face recognition for cameras (and law enforcement), neural nets for sensors, medical imaging and diagnosis, OCR … the list goes on.
Will generative AI fade? Most likely, it already is to some extend, but the Transformer technology, and the latent diffusion models are not going away.
3 Likes
For the past year and a half, I’ve been engaging in a bit of performance art on Facebook by using generative AI to write comments in the voice of a pretentious reviewer. And, well, maybe I’m just easily amused, but I’ve found that its output regularly makes me laugh out loud.
2 Likes
I know. Sometimes AI comes out with some nuggets. Usually (mostly) the outputs have to be edited down and adjusted, but it’s still a thing.
I’m realizing that it’s too much of a sweeping generalization to reject it all as simply not worth reading. Unedited output yes! But AI text used as part of the creative workflow process definitely has a place.
1 Like