Consolidated AI Thread: A Discussion For Everything AI

I’ve been using Grammarly pro for years as a slightly more advanced spell and grammar checker, but I’ve never liked their “rewrite suggestions” since they remove my favorite ways to abuse the English language. It’s really useful because you have control over what things you do want to use in it, but with their push to AI, I am unsure if I will renew my yearly subscription come October. On the other hand, I wonder how much of it is just jumping on the marketing bandwagon… it’s not like it generates text for you, just suggests ways of changing the text you have already written to fit a more general view of what English is supposed to do depending on the “style” you pick.

There might be an option to generate text in there somewhere, I just shut down all things like that on reflex. Might have a look around when I have time.

23 Likes

I bet there’s a lot of this going on.

“AI” is more of a marketing term than an academic one. It can refer to a plethora of algorithms, some that don’t even have anything to do with machine learning or generative algorithms.

19 Likes

You know what, it’s potentially worse than that. My university unveiled their new AI chatbot/checker (I think they’re worried about people putting sensitive research into places like chatGPT and having that spat out into someone else’s random enquiry if I had to make a wild guess- but I am guessing.)

I took a look at it and figured I’d try it out with a question that someone else had recently disputed me on that I had been meaning to do a literature search but hadn’t had the time yet. So it spat out an answer that was what I expect. Cool.

“Ok chatbot, can you please tell me what the references are for this statement?”

“Sorry human. I think you’ve misunderstood my limitations. I can only give general information, I do not have any references for that statement”

“Hmm, ok. Where can I find information on question asked to support this statement”

I don’t have any references for precise statements human. I suggest you look at some journals in this area or maybe read this very general textbook on the subject that may or may not have what you’re looking for…

Sooo, lets just think about this for a second. The AI is giving me information, but it has no idea where that information has come from, and it has no capability to find it. There’s no way to track down the source, work out copyrights or even if it is a credible piece of information. Given how much garbage probably gets scanned by AIs, and even how much information it has access to is now outdated, if it can’t tell you where it is sourcing the statements it is making, how can you ever be sure it is actually telling you the correct thing? I get it probably tries to work out what seems most current and common, but that is going to surely be very prone to error. If you don’t know what you don’t know, you can’t know when something isn’t right.

I had to laugh, I was speaking to someone recently and AI came up and I was saying how badly AI sometimes mangles summaries of journal articles etc if it misunderstands them and comes up with some very “interesting” statements of fact. They agreed and referred to it as the AI “hallucinating”. Now every time I see an AI going rogue I image it having some sort of fever dream.

Anyway, the issue is actually broader than just having AI make assignments and books. It’s being used everywhere. For example it was causing microsoft issues after it sounds like they fired most (all?) of their human editors which previously managed their news site: https://edition.cnn.com/2023/11/02/tech/microsoft-ai-news/index.html

Have you forgotten the bing chatbot so quickly? :joy:

^This. There are all sorts of CC’s that cover exactly how you may use a work from completely restricted to completely free and everything in between. For example you could have one under free but ND which means you can’t alter it. Or free but SA, meaning you HAVE to let others use what you create with it (share alike). Or you can zone it non-commercial if you only want people to make your work available for free. Just because something is “free” does not mean you lose control over how it may be used unless you specifically zone it to the public domain. If it is in a demonstration, unless you zone it otherwise (or you use material falling under one of the categories I’ve just listed like SA) you retain full copyright to it.

11 Likes

I concur. The issue with AI generated text is precisely that it is hard to detect. Because there are only a few real cintext specific hints that can get you an idea. Like a plot that constantly contradicts itself or events that start and finish in the same paragraph that come from the way prompts are read or the way an AI tends to limit how much it generates.

But I will always oppose this idea that you can detect AI by the prose of writing, the words use, or the vibe given by the text. There are billions of english speakers. From all walks of life from all corners of this earth, their english will undoubtedly be affected by their native culture and languages. Ranging from how they say words to how easy it is to grasp certain concepts relative to the concept in their native language. Be it the concept of warm harbors which really only exists in a country with unique geography like Russia. Or the habit of Japanese speakers to pronounce an L like an R due to the lack of an L sound native to japanese letters. These are all differences that will affect the writing, the wordage and the prose, let alone what you grew up watching that will affect your preferred style of english. Be it westerns or james bond movies.

It can be very insulting too to say that someone’s writing is AI. Since you are basically invalidating all the work that went into it. And at worst, calling them a leech. So just because someone has a unique style of speech or prose. Irrespective of if its something that can be seen as similar to the style chatgpt uses. Is not in any way shape or form evidence of AI usage.

16 Likes

There’s a huge difference between spelling- and grammar-checking tools (which may call themselves AI as a marketing gimmick) and generative AI. The former takes a sentence you’ve already written, compares it to its database, and makes suggestions to help your content conform to what it knows. Generative AI claims to write “original” sentences, paragraphs, chapters, whole finished works of art, from just a few words of prompting, but really is just mashing together words and strings of words (or images) that are statistically likely to be found together in the millions of works it’s scraped. Basically, Gen AI is a glorified autocorrect, “predicting” which words will come next, and we’ve all seen how absurd and meaningless autocorrect-generated sentences can be.

Ironically, neither of these are actual Intelligence (Artificial or otherwise). These programs are not sentient and do not understand or think about the content that goes in or out. It’s content that might briefly look kinda sorta like what a good piece of original art might look like, but usually falls apart under any close examination.

ETA: Whether or not you want to keep your Grammarly subscription is up to you, but even if the use of the word “AI” is just a gimmick, I find it disturbing that they’re jumping on that bandwagon, and I’d be concerned about it scraping my words for whatever gen-AI project it’s got cooking.

22 Likes

Pixels, more likely, but anyway.

I’d personally call it a glorified chatbot, but I’m not sure that’s any clearer (especially if you don’t have any history with chatbots :laughing:)

(I still want to write that Markov chain algorithm for my conlanging project though.)

This is precisely why, even though I have Grammarly keyboard installed, I never use it. I don’t want my random typings to be sent anywhere.

4 Likes

it’s technology, just like in any art discipline. That will stick or move into the moat.

I don’t think this technology work by stealing work of people more like copying works of people. And mixing it a bit to make look like its a new thing.

Like any tool in compute line, people will use it or reject it. Either way the thing that matter is
" Does that make good games or good storys ? ".

1 Like

That’s, like, plagiarism 101.

17 Likes

Its basically all anyone does. Unless you’re some kind of blind deaf person who came out of the woods and wrote something wholly original in a language they made up. We take in what has come before us, mix it up, spit it out.

Edit: I think an extended discussion focused on the inherent superiority and divinity of human thought is better left to the ai thread than this one.

2 Likes

No it is not I don’t think anyone here copies one by one literally phrases from 300 books copying literally plots from copyrighted material vomiting random words without any craft or intent.

I won’t let people say literally we all here are a group of con artist without talent stealling other works

1 Like

Also there’s a difference between drawing from reference picture and tracing the reference picture.

10 Likes

People here totally will say Picasso was an Ai that copy Velazquez. The juggling mental games some can play to not be able to admit that using an Ai to writing random stuff is not writing it is at best producing or edit others talent butchered by a meat grinder

3 Likes

Yeah, we might bring all of our myriad experiences with us when we create something, but we actually think about what we create and make artistic decisions within the context of our lives and identities. A program might roughly mimic that process with all the data it’s ingested, but it’s not actually doing that. It can’t truly do that, no matter how detailed and nuanced someone might think their prompt to be.

You’re better off trying to make your own thing with that same prompt; that would be the most honest creation, no matter how lovely the AI output might appear in comparison to an unpracticed hand.

10 Likes

If that were literally true and there is no originality in any art and we are just copy each other there wouldn’t have at all any advance on any art would still in the caves.

Negating culture advance and originality of humanity is negate all we are and we were to reach this points.

But sure we are all a brainless program picking random lines to fill a math formula.

4 Likes

This makes me think of how so much of humanity’s growth has been a collective, community effort. We build our knowledge with what others have learned and bring something of ourselves to it and so create something new to share, but I don’t think generative AI fosters this sort of growth.

It feels anti-community to me, when someone chooses to generate images or writing with AI for a published project instead of collaborating with others (especially when we can connect to so many people with the internet).

8 Likes

Yeah? Thats why so many artists and writers are suing the AI companies and the courts are willing to listen. Like others said, its plagiarism.
Plagiarism alone would keep a lot of lawyers in champagne for the rest of their lives. Its a very profitable part of the legal profession, and it wouldn’t be if plagiarism and stealing the work of others were legal. There are dozens of Youtube videos talking about the people who did this and got sued. Successfully. Even if it was accidental.

1 Like

This is literally the definition of plagiarism.

10 Likes

Anybody who thinks you can identify AI writing because “It uses language that’s too flowery and reads weird” has obviously never had to read papers from freshman comp classes.

9 Likes

I am extremely glad that CoG/HG is taking a hard stance against AI-generated bullshit and I will be reporting things (not just here, but in other game communities where some of their games are using AI resources as well) that way the proper people are made aware.

There is far more to it than just the blatant overuse of “amidst” and “embarked.” The prose generated by AI isn’t just purple; it’s pointless, repetitive, meandering, disjointed in numerous places and hits a brick wall (or let’s say full stop) before jarringly switching subjects. Descriptions of things make little sense, as they sound pretty enough, as they go far beyond thesaurus humping, use metaphors that don’t make sense in any language (I am multi-lingual, I speak English, Japanese and German fluently, for what little that might be worth) and never really get to the point of explaining why the minute details of starlight gleaming off the mercenary woman’s poorly-described weapon matters, as that detail is forgotten as quickly as it’s generated. Issues that it generates as “plot points” are resolved strangely quickly and there are no repercussions.

Those are just a handful of indications that there’s an extremely high likelihood that the writing is AI-generated. Then, while stilted dialogue can and does happen even when an extremely skilled author is writing it, AI-generated dialogue makes the characters sound like they’re spitting surprisingly stilted nonsense (and occasionally verbal prose as well.)

Honestly, it’s extremely easy to differentiate between something an ESL author, or someone who’s been having a hard time writing for whatever reason, has come up with-- and AI-generated garbage. People like to disagree with me for… well, reasons, basically, but once you’ve read more than one AI-generated story, you’ll notice things that will stand out, regardless of what bot they use to spew their trash with.

It’s not really going on a witch hunt or alienating a group of people, when ESL writers’ works are usually pretty damn decent-- and are far better than what AI generates and throws in the metaphoric hog trough. It might take reading a few lengthy generated stories, but after two or three, it’s very evident.

2 Likes

I’m of the general opinion that any AI-derived work should not be saleable at any value, should not be eligible for copyright or trademark, and should come with something that makes it clear that AI was used in the process of its creation, ideally noting how and to what extent. But I believe that it should be perfectly valid to circulate and use with those stipulations adhered to.

I recognize that there are several problems with my opinion in execution.

As AI advances, such works will be more and more difficult to detect with certainty. It is already difficult to detect and prove the AI origin of any work, especially if it there is a mixture of some AI and some human-origin content. There’s also the question of the context of the AI-generated work, is it a fiction novel? Character art? A working program with actual utility? How much AI influence is enough to negate the value of the end product? Does having a bounty of freely available AI works smother interest in human works?

If an AI provides a pseudocode blueprint of your program, and then you write all the code based on that blueprint yourself, should you not be allowed to make profit off the sale of the program?

If an AI actually code the entire working program, one that works perfectly for its task, should you not be allowed to make profit off the sale of the program?

If an AI gives you a framework for a story, and you write the actual story following the framework and adjusting as you go, does that much assistance negate the labor you performed?

If the end user cannot tell the difference between the two, does it matter?

There are endless questions like these that make an easy answer difficult to arrive at.

All that said, I divide things into two camps, personally. Utility, sentimentality. In the utility camp, I’m looking at things like user interfaces/character/setting/iconography art, actual code, translations, editing, etc. If I want something for its utility, I don’t care about the creator or process behind it, it has a job to do, and it either does it or doesn’t. In the sentimentality camp, that’s where self-expression has value and the creator matters to me. A song, or poem, or painting that shows me how the painter views the world or some part or piece of it. A book that allows me to connect with someone, even a fictional character, on a human level.

My hope is that we can figure out how to use AI and advancements like it to make utility-oriented works widely, easily, and freely available, without smothering works with sentimental value.

I don’t believe in capitalism as an ideal of any kind, and wish we lived in a world where people’s livelihoods didn’t depend on turning themselves and their works into commodities for sale just to have a place to eat or sleep, and I think all of this would be easier in that world, but in the world we live in, I recognize that this is displacing jobs and talented people, and I regret that. I think we’ll survive it though. It will find its niche, people will adapt to its existence, we’ll move on, just like we always have.

4 Likes