I’m glad you had a positive experience with the welfare system, and while I’m unfamiliar with Finland’s, I’m sure that positive experience is not universal. It really depends on personal circumstances. And I don’t want to veer it too much into the realm of politics either. But it’s important to acknowledge that most Western countries run on a narrative hostile to the welfare system recipients.
I’ll admit, I was curious about what AI would do if you gave it a sample of writing to finish so gave it some of mine and kept prompting it to keep it on track. It was still very predictable, right to the point and kinda boring/bland. It didn’t seem to understand how to build up a scene with tension and just kinda went ok well this has to happen so here it is in a brief straight forward way, all done now. It’s not written in a way I would have used… But it did manage better than I expected.
Honestly I think it’s going to be a losing battle to completely keep AI out of creative areas if it continues to improve at this rate. I suspect writers are still going to be needed in some capacity to guide and refine anything AI writes but I can see the bigger corps like Disney liking it especially for their derivative stuff (remakes, sequels, prequels, spin offs) to reduce the writing team they need to hire. I do worry it’ll reduce anything truly original and creative coming out with use though over time the more it is relied upon.
I’ve already commented earlier I think that I find it rough when they used copyrighted works to train AI to make work “in the style” of existing writers/artists without permission or compensation. I’m honestly not entirely sure how it is legal? I’m guessing there must be some kind of fair use clause in there somewhere I guess?
Did you guys know this was a thing?
I work in technology for a multimedia company, I am extremely familiar with machine learning tools, I got started toying around with various art tech about three years ago, stuff like Gigapixel, artline, disco diffusion. Nowadays I am quite familiar with stable diffusion, less so with Midjourney, and I’ve never touched any OpenAI products. I have also used Voice.ai a bit, and use Tabnine all the time for work.
How often do you use AI art generators?
I like D&D but tend to make way more characters than I’ll ever play, and I have fun messing around trying to visualize them with SD. I’ve made my own LoRAs for the purpose. However, anytime I have a character I’m going to actually play (rare these days), I tend to commission one of a couple people whose art styles I like who don’t charge much for some line sketches, that I can use as tokens. None of the outputs I’ve generated ever get seen by anyone but me.
Do you think the Interactive Fiction industry and community will be changed by this? Think Tumblr, Discord, and other social media. Will this be more of a topic going forward as AI-generated art becomes more popular?
Yes, LLMs for art generation are cool but ultimately an evil that will do more harm than good. emad and others can talk a big game about wanting to democratize creativity, to try to work towards Star Trek holosuites or whatever, but ultimately all of the current paths are going to dead end fairly soon because the models simply do not have comprehension, they are at the end of the day just (very very advanced) statistics. They are going to plateau right around the point where they’re extremely useful for fucking over writers, actors, artists, and similar - but this path doesn’t take us to GAI.
How would you feel if CoG decided to use AI art?
I’d be furious. These tools can be fun to play around with, I won’t deny that. But trying to use (most of) them for profit is deeply unethical. They’re better defined as plagiarism tools than AI. I left academia a decade ago but I worry a lot about what’s going to happen to a lot of the social sciences and education generally as people who don’t understand these things start to use them to replace learning and understanding for themselves.
Do you think AI generated art is going to be more beneficial or harmful to your reading experience? In more focus, would you be okay with it if it helps create more art in your games, or do you rather want to use your imagination?
I think a day might come when there are tools like KoboldAI and AIDungeon that are significantly more advanced and capable of creating mostly-consistent worlds for stories to occur in, enabling players almost complete freedom to act within those worlds. E.g. feed all the various books, movies, comics, shows, and games set in the Star Wars universe into a future LLM as a LoRA and tell it you want to play KotOR 3. That will likely be possible to some extent in the next ten years or so. I think in the next five years we’ll be able to recast roles in movies however we want. If you like The Usual Suspects but feel icky about Kevin Spacey? Point a tool towards the mp4 file and towards a model built on images of Will Smith, let it process, now you watch the movie with Will Smith as Keyser Soze. That is already possible though limited in terms of quality -and that is exactly the kind of thing that studios want to do, and all the guilds are striking against and they are 100% right to oppose it. Deepfakes are already very good, they will get better, and we all have to take a hard stance that they cannot ever be allowed to be monetized. Consent of the talent is another major problem here, but the technology exists and will be improved, so we have to accept that as a reality and find ways to protect people. I honestly don’t know what the answers are, but it’s important for all of us to try to fight against anyone profiting from LLMs that have been trained on stolen work.
With regard to CoG and text-based interactive fiction, it’s likely that in a few years it will be possible to feed an entire game into an ML tool and use it to turn a text game into a fully-animated 2d or 3d experience. I think that’s coming and I think it has to be something hobbyists can do for fun - but if corporations try it they need to get sued into oblivion by the artists whose work went into the model without their consent.
I see AI art generation as a little bit like Uber and Lyft. It’s really neat until you think about it and realize the true purpose is to destroy unionized workforces.
And if those unionized work forces consistently underperform and refuse to innovate?
One thing i have recently been quite successful with, are the current AI story telling algorithms; they have taught me how to write.
English isn’t my native language, nor do i have any background in writing what-so-ever.
But something these algorithms such as NovelAI have allowed me to do is to learn from an example.
I would copy&paste a few pages worth of dialogue between two characters written by someone far more talented than me(To have it as the baseline example for the algorithm), and then i would use the AI to generate lines based on that, and i would go back and forth with the AI, each of us typing a few lines in turns.
At first my input was very stiff; i only wrote direct quotes or very simple description and didn’t elaborate beyond what was absolutely necessary, and very quickly the AI picked up on that and the entire story went into this death spiral,where it adopted my poor writing.
But eventually i learned to write stuff like: “This is horrific!” said Werumenz as he discovered the horrors of writing, fear in his eyes.
Yes, very basic stuff, but to me actually going into more detail to describe just the basic stuff happening around the scene or with the characters i had imagined was kind of a breakthtough, and the algorithm picked up on that. In the past it only took a page or two before my stories ate themselves by becoming these spiraling dialogue sections about things that didn’t matter to anyone in the story. But once i discovered the esoteric knowledge of describing character’s thoughts and the overall scenario in general from the third perspective, the quality of my AI generated stories improved immensely.
What i love about AI assisted writing is that i have immediate feedback, as the algorithm learns from my style. So if i write well, the feedback will be well written lines, but if the generated text feels bad, there probably is a way to fix it by improving my earlier written descriptions of situations, characters and/or their motives or something.
This is at least my experience as a very amateur writer.
Maybe this isn’t the proper way to learn how to write, but if it weren’t for these algorithms, i wouldn’t be a writer at all. So these things have at least to some value to someone.
What’s your thoughts on ai and people using them make stories(writer’s block). And do you use ai yourself? I have seen some threads that talk about using AI tools for creating stories and scenes, nevertheless.
Do you use AI tools for writing? If so, which ones and how do you use them? If not, why not and what do you think to what extent should it be used?
I mean there’s AI tools I use occasionally, mostly for generating ideas or inspiration. I have tried some tools like ChatGpt, Google Bing, Google Bard, AI Dungeon etc
I think they are impressive in some aspects, but also flawed in others. Sometimes they can produce coherent and engaging stories, but other times they can be inconsistent, repetitive, or nonsensical. So we have a long way till ai replaces us. As of now they, they still need human guidance and editing to make them polished.
And then there’s Ai Art…
AI will lead to the downfall of humanity, in my opinion.
Basically resistance is futile. If the US or EU countries don’t embrace it then countries like PRC will. I don’t think it will be that long before most “human only” work will be done by or in conjunction with machines (which will still dramatically reduce labor demand). There is a bit of a perverse incentive for business because labor is always an overhead expense while machines are an asset. So even if the improvement is lateral or even slightly suboptimal (self-checkout for example) most business will still convert to it. I think something like UBI will be necessary to mitigate the massive wealth and power imbalances automation and AI will bring.
I’m optimistic about it, but still have reservations/worries, but more about the way people will use it as a tool and because of societal pushback/unrest. AI as a tool is here to stay, it won’t go anywhere. If western countries don’t keep up with the game and let hostile actors get a big lead, it can end up badly - imagine TikTok’s algorithm debate but on steroids.
As a scientific tool, it has gigantic potential. I follow a couple of science podcasts and I think almost every scientist I’ve heard talking about AI has been extremely excited over AI and all the ways it might further their fields - astrophysics, SETI, biology, medicine, electronics, etc.
I personally don’t see possible AGI as that different from humans, or a “mind/intelligence”, just because some existing people are terrible doesn’t mean we should stop, well, making humans because some of the new humans miiight be terrible beings. I’m also not sure if AGI as we define it today is possible, we might have to reevaluate what it entails since we have absolutely no idea how differently (or similar!) they might think or see things compared to us.
Regarding the subject AI in art, I think the issue gets a bit overblown. For basically all of time it was thought that art was going to be one of the last fields touched by AI, turns out it’s one of the first fields, so of course it generates questions and drama. I think some regulation regarding copyrights and training sets will be needed (although, personally, I don’t see thaaat much of a different between a little kid going to an art museum versus an AI looking at it, but I also understand why some people do see a difference) so that artists can be sure that stuff uncannily similar to their work might not start popping up around and making money for others or earning accolades. One thing I’m very, very opposed, thou, is “reviving” dead actors, singers, painters, etc. with AI to “finish” a song, movie, whatever, unless the person explicitly gave permission in life.
Regarding the economic impact of AI, I’m still formulating my opinions and need to read and think more about it.
Edit: As a writing tool! I’m not a writer, but if I was, I wouldn’t use it except to clean up typos or do some kind of proofreading. I think it can also be a great learning tool for people who are writing in a language different to their first language, since structuring sentences and ordering words can be a challenge for many.
I have long used AI Dungeon for personal entertainment purposes, but I wouldn’t feel okay using it for anything I wanted to actually publish. I do find it really helpful in combating writer’s block.
Recently, I have started considering whether I want to use it at all, since a lot of the same ethical concerns that we see with visual art, also apply to the writing AI’s.
I did try using ChatGPT for brainstorming story ideas, but found it’s ability to generate and elaborate on ideas both shallow and very stereotypical.
Generally, I feel the same way that I do in regards to visual art:
If the data is sourced ethically, and the AI is used as just a tool, as a part of the process, instead of just replacing it, then I think it could be useful.