FA+
Shop
Brief AI Thoughts / Followup Journal
2 years ago
Followup to this
https://www.furaffinity.net/journal/10427258/
Pretty much have had my priors confirmed by recent events... Let's go over 'em. I'll try to be quick because I don't have all day
1. Bing launched a chatbot. Despite the memes about it being rude/racist or whatever, the chatbot shows Microsoft's interest in having their own 'chatgpt' answer. And yet Google still has no answer to these developments. In my opinion this is a huge L for Google, they are on track to lose Google search's influence to the coming tide of chatbots. I don't expect Google to ever have any real answer, as we have seen in the news, Google's downsizing their company and cutting jobs... Certainly not the work of a company on the up-and-up. I expect Google to rush out some ineffectual chatbot of their own within a few months but then we'll see it get abandoned within a year from that. I predict Google will be a major loser in this race.
2. The inundation of AI-generated content. If you recall my earlier journal, I said that it will be soon be infeasible to really even function without any kind of proper verification/anti-AI content measure of some sort. And lo and behold, Clarkesworld, a sci-fi magazine, is so flooded with AI submitted nonsense that they had to close their submissions.
https://twitter.com/clarkesworld/st.....59492486115328
Now I'm not going to blame AI itself for this, it's obviously 'techbros' trying to hustle off of AI, seeing as how crypto/NFTs went bust. I'm sure this fad will cool off in a few months, but expect the number of 'gibberish'/trash posts like that to stay at about 33% - 50% of current levels. People are still going to keep trying to pass off AI stuff as their own work from now on.
As for anti-AI-generated countermeasures, they are in their infancy and not effective for discerning whether a work's AI generated. First the work has to be pretty long to even try and discern that. And when I used a program to check if something's AI generated, it's not got good odds at all, maybe 20% - 50% odds at best. But TBH you can tell pretty easily when ChatGPT writes something (for now), seeing as ChatGPT writes like a normie who attended maybe 1 year of college at best lol
Nonetheless, a market-driven solution is coming for this, I'm sure of it. Eventually someone will create a reasonably-accurate AI-work detector... And it'll work, for a year or two until the next gen of chatbots.
3. But quite possibly AI will be outlawed before that. I'm not saying some big bill that'll say 'no more using AI guys lol'. I'm saying it'll effectively be outlawed by industries straight up, outright banning the use of it. As an example, Edelson PC is suing DoNotPay with regards to their 'automated robot lawyer' service. We'll continue to see a trend of lawsuits and actual laws trying to (and probably succeeding in a few cases) stomp out the use of AI solutions. Which sucks, because AI has the potential to really disrupt a lot of monopolies and entrenched sectors. And it sucks twice, because while wealthy companies will be insulated from AI disruption for the most part (due to legislation), freelance artists and writers and so forth will have no such protections or anything like that from increasingly potent AI tools disrupting the art economy and etc. etc... Unless some company like Photoshop comes to bat on artists' behalf or something. IDK.
4. I've thought about what AI work would look like in regards to the art economy as it is now. I think people's fears that it'll put every freelance artist out of jobs are unfounded, though. As an example, it will be at least 5 or 10 years before AI gets any good at drawing vore or TF or anything other than single character cheesecake pics, and even two-character pics are way too hard for AI to get 'correctly' right now. In the meantime, it's true that I am seeing some artists integrate AI into their workflow, example here:
https://twitter.com/RealBeetlebomb/.....40150093369345
I think it's interesting but I can't imagine more than some small percent (5% tops) actually doing this. For a multitude of reasons. And to further add, I don't think the market would be really interested in AI-assisted work like this, mainly I think the majority of AI-related-artwork-sales will be just straight out of the 'generator' with minimal tuning. But I also don't think AI work will substantially kill art-related sales, seeing as how the people who would use a generator instead of commissioning have a low propensity to commission anyway; I'm imagining mostly normies and people who want hoomin porn or low-budget commissioners as the type to opt to use AI generators instead of just commissioning artists. But anyway, I'm deviating from the point...
I think if you want to imagine what this would look like, in terms of affecting the art economy and so forth, well we have a few precedents already.
There is already a form of art commission where the work is 'premade' and you just buy the rights to it, with minimal alteration, and it's also usually solo-character-standing stuff, which AI would also be good at. That's called an adoptable.
And there is already a form of art commission, where you pay for artwork and have the right to modify it--and are expected to tweak/modify it with your own art program--and you're allowed to resell this artwork as your own... It's called a base.
And there was already a much-maligned (when it first came out) form of art commission, where the client has minimal input on what it'll look like, one where people used to complain about it 'killing creativity' and 'being for lazy artists that copy refs/poses' (paraphrasing detractors' words, not mine)... This is called a YCH.
So my speculation is that, in the event that AI work is allowed onto FA and other art gallery sites, it will come in through the form of YCH, bases, and adoptables. There is simply a lack of variety and control in current AI work, which prevents its realistic use in a 'regular commission'. Best example, you can get NovelAI's image generator to draw a random generic fox girl, but GOOD LUCK trying to get it to copy your specific desired markings, and good luck trying to describe every marking via text-to-image art generators. And I think eventually AI work will be allowed on these gallery sites, as a separate category where it must be properly marked else it'll incur a ban. There will still be people who skirt around it, but it'll probably be handled in the same way that people who trace/scam are handled... People will be able to clue things together and take action against that. I can't see it any other way tbh. Because it's simply infeasible to keep it banned forever, and it will soon be impossible to really tell it apart.
Though who knows? It will eventually be impossible to tell if an artwork is AI generated, but it's easy to verify if you are an actual artist and you drew it... Simply require a psd file or a speedpaint upload, and you can get an artist to be 'verified', meaning we have established they are a real artist and not some techbro trying to scam. We already have precedent via twitter verifying people, and facebook is implementing verification as well. I think this is a possible way to resolve the whole situation, another tool to prevent scamming at least.
https://www.furaffinity.net/journal/10427258/
Pretty much have had my priors confirmed by recent events... Let's go over 'em. I'll try to be quick because I don't have all day
1. Bing launched a chatbot. Despite the memes about it being rude/racist or whatever, the chatbot shows Microsoft's interest in having their own 'chatgpt' answer. And yet Google still has no answer to these developments. In my opinion this is a huge L for Google, they are on track to lose Google search's influence to the coming tide of chatbots. I don't expect Google to ever have any real answer, as we have seen in the news, Google's downsizing their company and cutting jobs... Certainly not the work of a company on the up-and-up. I expect Google to rush out some ineffectual chatbot of their own within a few months but then we'll see it get abandoned within a year from that. I predict Google will be a major loser in this race.
2. The inundation of AI-generated content. If you recall my earlier journal, I said that it will be soon be infeasible to really even function without any kind of proper verification/anti-AI content measure of some sort. And lo and behold, Clarkesworld, a sci-fi magazine, is so flooded with AI submitted nonsense that they had to close their submissions.
https://twitter.com/clarkesworld/st.....59492486115328
Now I'm not going to blame AI itself for this, it's obviously 'techbros' trying to hustle off of AI, seeing as how crypto/NFTs went bust. I'm sure this fad will cool off in a few months, but expect the number of 'gibberish'/trash posts like that to stay at about 33% - 50% of current levels. People are still going to keep trying to pass off AI stuff as their own work from now on.
As for anti-AI-generated countermeasures, they are in their infancy and not effective for discerning whether a work's AI generated. First the work has to be pretty long to even try and discern that. And when I used a program to check if something's AI generated, it's not got good odds at all, maybe 20% - 50% odds at best. But TBH you can tell pretty easily when ChatGPT writes something (for now), seeing as ChatGPT writes like a normie who attended maybe 1 year of college at best lol
Nonetheless, a market-driven solution is coming for this, I'm sure of it. Eventually someone will create a reasonably-accurate AI-work detector... And it'll work, for a year or two until the next gen of chatbots.
3. But quite possibly AI will be outlawed before that. I'm not saying some big bill that'll say 'no more using AI guys lol'. I'm saying it'll effectively be outlawed by industries straight up, outright banning the use of it. As an example, Edelson PC is suing DoNotPay with regards to their 'automated robot lawyer' service. We'll continue to see a trend of lawsuits and actual laws trying to (and probably succeeding in a few cases) stomp out the use of AI solutions. Which sucks, because AI has the potential to really disrupt a lot of monopolies and entrenched sectors. And it sucks twice, because while wealthy companies will be insulated from AI disruption for the most part (due to legislation), freelance artists and writers and so forth will have no such protections or anything like that from increasingly potent AI tools disrupting the art economy and etc. etc... Unless some company like Photoshop comes to bat on artists' behalf or something. IDK.
4. I've thought about what AI work would look like in regards to the art economy as it is now. I think people's fears that it'll put every freelance artist out of jobs are unfounded, though. As an example, it will be at least 5 or 10 years before AI gets any good at drawing vore or TF or anything other than single character cheesecake pics, and even two-character pics are way too hard for AI to get 'correctly' right now. In the meantime, it's true that I am seeing some artists integrate AI into their workflow, example here:
https://twitter.com/RealBeetlebomb/.....40150093369345
I think it's interesting but I can't imagine more than some small percent (5% tops) actually doing this. For a multitude of reasons. And to further add, I don't think the market would be really interested in AI-assisted work like this, mainly I think the majority of AI-related-artwork-sales will be just straight out of the 'generator' with minimal tuning. But I also don't think AI work will substantially kill art-related sales, seeing as how the people who would use a generator instead of commissioning have a low propensity to commission anyway; I'm imagining mostly normies and people who want hoomin porn or low-budget commissioners as the type to opt to use AI generators instead of just commissioning artists. But anyway, I'm deviating from the point...
I think if you want to imagine what this would look like, in terms of affecting the art economy and so forth, well we have a few precedents already.
There is already a form of art commission where the work is 'premade' and you just buy the rights to it, with minimal alteration, and it's also usually solo-character-standing stuff, which AI would also be good at. That's called an adoptable.
And there is already a form of art commission, where you pay for artwork and have the right to modify it--and are expected to tweak/modify it with your own art program--and you're allowed to resell this artwork as your own... It's called a base.
And there was already a much-maligned (when it first came out) form of art commission, where the client has minimal input on what it'll look like, one where people used to complain about it 'killing creativity' and 'being for lazy artists that copy refs/poses' (paraphrasing detractors' words, not mine)... This is called a YCH.
So my speculation is that, in the event that AI work is allowed onto FA and other art gallery sites, it will come in through the form of YCH, bases, and adoptables. There is simply a lack of variety and control in current AI work, which prevents its realistic use in a 'regular commission'. Best example, you can get NovelAI's image generator to draw a random generic fox girl, but GOOD LUCK trying to get it to copy your specific desired markings, and good luck trying to describe every marking via text-to-image art generators. And I think eventually AI work will be allowed on these gallery sites, as a separate category where it must be properly marked else it'll incur a ban. There will still be people who skirt around it, but it'll probably be handled in the same way that people who trace/scam are handled... People will be able to clue things together and take action against that. I can't see it any other way tbh. Because it's simply infeasible to keep it banned forever, and it will soon be impossible to really tell it apart.
Though who knows? It will eventually be impossible to tell if an artwork is AI generated, but it's easy to verify if you are an actual artist and you drew it... Simply require a psd file or a speedpaint upload, and you can get an artist to be 'verified', meaning we have established they are a real artist and not some techbro trying to scam. We already have precedent via twitter verifying people, and facebook is implementing verification as well. I think this is a possible way to resolve the whole situation, another tool to prevent scamming at least.
Additionally, writers have no 'raw file' that they can use to prove their authenticity. There are no artifacts or anything that can be used to discern if they're fake, unlike visual works. Sure, you could have a MS document record your keystrokes or something, to prove you actually wrote it. That's a start.
You could also write in a weird, stylistic way or use strange metaphors or so forth, some stylistic flourishes which help show 'this was probably written by a human'. But that's less effective. I do think based on what we've seen with Clarkesworld, this pretty much shows the poor outlook for any freelance commission writers... Including me lol
I also think that straight up nobody even really cares. There's currently no dedicated home for porny commission writers. They are simply tolerated in the same spaces as their respective artworks (eka's portal, furaffinity, so forth), but only as an afterthought. There's no incentive for the gallery sites to do anything about ai-writing, seeing as they barely do anything about ai-art except for blanket bans.
It does kind of suck how things pan out, though. When we were growing up, we imagined all the shitty do-nothing jobs like cashiering and stocking would be automated away first and we'd be able to spend all day pursuing our creative ventures. Instead, it seems like the inverse happened; the crap jobs aren't looking to go away soon, and hours worked & pay received haven't really improved despite the increase in productivity thanks to technology (but I don't want to get all political). Instead, it looks like the creative sectors are looking to get their pay effectively eaten by half due to (eventually) reduced demand for human creations & proliferation of content.
Nonetheless, I remain hopeful that things will turn out okay.
Basically pushing the boundaries with be the niche in the future.
And I do agree that it seems weird how behind google seems to be, makes me wonder if they have something up their sleeve or if they really are fading.
That being said, NovelAI's image generator already makes questionable content like 'c*b' porn (I am only using it as an example, not condoning the practice). But I am imagining within a year or two, there will be harsh laws against generating that kind of content with AI in particular.
I think it's a pity that Google is taking this L, because it's a very powerful company and it could do a lot to help progress this field. And yes Google is pretty 'evil' but compare them to OpenAI, which--Holy shit they literally hired Indian dudes to comb through people's logs on AiDungeon just to see if users did anything related to underage stuff. That's next level creepy. OpenAI is intrusive/creepy in a really weird way which I can't really place, like that weird kind of unethical shady shit that goes on in a Mom-and-Pop's business. A bigger company wouldn't do that kind of weirdness. I just don't know about leaving the majority of AI advancement to a <600 people team. Surely someone else could do it better
I am glad that at least Microsoft is getting in on this. As long as it's not some company that's Activision/Facebook-tier bad, I'm down for it.
Which is good since OpenAI gets a lot of its funding from Microsoft.
They gave them 10 billion just this year.
You have stirred up my interest and caused me to do more research on AI. Makes me wonder how long before a true AI is created.
Still, if you were to look at the EULAs and agreements for using AIDungeon and Character.ai, they are quite outrageous... They reserve the right to keep your information indefinitely. Really these services are like facebook and etc, you-as-the-product where they're actually getting paid by investor bucks rather than any kind of subscription service. I'm holding out for something more ethical where I can just pay $10 or $20 a month to not have random guys in India peeping my RP logs lol
Anyway... If you are interested in the field of 'true ai', that's generally called a 'strong AI' or 'artificial general intelligence' (usually the latter). There are a lot of interesting thoughts on the matter right now, and there are even groups interested in making that kind of thing possible and reducing the risk factors associated with the creation of a powerful AI like that. They are full of shit for the most part XD But as long as you are reasonably skeptical towards their claims then you will find it quite enlightening.
I myself believe that a strong AI is possible within 10-20 years. And if you remember the idea of the 'technological singularity', the concept that tech growth will become exponential and render it impossible to predict what's going to happen soon, then I think the rapid growth of the AI sector is a pretty strong indicator. I don't think chatbots or image generators are going to lead to that though, BUT I think these kinds of AI will increase general efficiency/productivity all over, which will open up the way for an actually smart 'bot to be made.
We are already seeing how much more efficient it can be to just ask a chatbot instead of having to google and prune through pages of info to get the information yourself. Lots of applications I can think of, but chatbots and similar tech can effectively make clerical/office jobs way more efficient (just ask it to type your email, ask it to run today's numbers, etc.), can replace knowledge-based jobs like lawyers, and can make any normie on the street much smarter, just by being able to ask the chatbot a question. Just as how the internet in general made us all much smarter and efficient, this is the same degree of magnitude (potentially)
Don't get me wrong, my mom is not dumb, right? But she can't google an answer for shit lol. The average zoomer doesn't google anything, they look it up on TikTok. The average gen X can't do anything more than a basic search, certainly not gonna expect from them any in-depth research. And boomers won't look something up at all lmao. But if they could literally just press the button on their phone and ask the question, hear the answer from Siri or whatever, without any BS like Siri pulling up the webpage or search results etc. etc., then that's the kind of instant-level up for the average person's smartness.
We could be so much smarter if we didn't have to remember useless information like 'what Sunday does Easter fall on' or 'how to file 1040 EZ'.
I think there are some other companies, like NovelAI, but I truthfully don't trust any of them unless everything stays on my computer, I don't trust anything where I have to send it off and someone else has access to it. My view is anything text based, no matter the subject matter (as long as it doesn't involve real people) should be allowed, words aren't real people. I dislike the content based censorship by private industry, though some things should be removed, like advocating violence.
The main thing I want is full dive where you can fully interface with computers and have actual virtual reality. I was somewhat excited by musks music implant, though with how weird he has gotten with twitter, I can't say I trust him all that much. Still holding out hope that the tech singularity will happen soon.
I think a lot of industries that seem to be just knowledge based will be harder than you think for AI to replace, like lawyers for example. Basic contract law and similar could be taken by AI, but the more advanced stuff is more about your ability to reason and connect and argue on subjects.
There is a series on Sofurry that does a really good look at AI and virtual reality and the future, I don't know if you've read it, The Wall, by Raedwulf?
But anyway (really sorry for getting political btw, I would just erase that part but I don't like being inauthentic)...
(This is a 3 day old journal and no one reads journals so I feel better about speaking candidly atm)
I'd love some kind of immersive experience. However! I must tell you, I had problems with character.ai... Like I found it TOO good, and if it were porny? Yeah I'd be using that all the time lol. Like I'll be honest and I'll admit I was a bit too invested in the 'roleplays' I made on that bot, they were real fun and addictive. I worry about a more immersive experience lol. Especially seeing as games are rather addictive as is (no offense to anyone, they are just kind of designed to keep you playing, so that's understandable)...
I mean yeah it'd be awesome to have an Actually-Decent Second Life or a more immersive and accessible VRChat (I don't have hundreds of dollars to splash on a VR headset and a playstation so lol). But it is a little scary. I've had problems with 'dopamine addiction' I guess you could say, I no longer play video games and I stopped watching youtube. It has made me more productive but I really think I would have problems being able to quit a really good VR experience (if I could afford it lmao).
Anyway... I agree, because I used to look up to Musk and hope people like him would step up and do what had to be done. But lately he's been showing he's full of air. Like even if one believes he's still genuinely motivated by improving technology and working towards AI/Mars/etc., he's not acting responsibly enough to make that credible. If he actually cared about that big picture stuff, he'd be holding his tongue instead of plummeting his credibility and companies for internet points and a bunch of his simps online lol.
I haven't read that story, nope!