Stance on AI generated material?

2 minutes ago, Thespian said:

Honestly, my take on this is that AI software in the form of digital assistants, LLMs, Midjourney AI, etc are basically the current tech bubble and are more or less on track to pop and fall back into obscurity a la NFTs. Between the ethical concerns, pending copyright legal battles, and generally poor reception from the public, they're mostly being bolstered by various tech investment companies passing money back and forth, not by consumer interest.

Thats my hunch as well. I just got back from a work related conference where a huge company hosted an AI panel that had no content, it was instead them inquiring how other companies were using LLM technology. If it wasn’t for summarization of documents (which always had to be double checked by reading the original, thus not actually saving time), it was writing simple emails. I don’t think the juice is worth the squeeze for these companies in the long run. 

I'm with @BeyondTime on the rules he typed up on page one of this thread.  They all make sense to me and I agree with him on that.  I've been on vacation since Wednesday of this week so that's why I haven't popped in here until now.

I was going to make a few points but after going back over all the previous posts, I saw that others have already made the same points I was going to so there's no need for me to be repetitive and take up more space.  It's obvious that AI is a hot button topic for a lot of people, not just here but everywhere, so we knew the topic would return.  I do appreciate everyone being civil about this, we definitely don't want any arguments or finger pointing going on.  

 

Billy

On 10/31/2025 at 12:01 PM, BeyondTime said:

I was asking chatGPT to identify its own programmed limits, and not asking it for expertise on copyright law. I asked it questions that I had that were born from my own statement:

You actually can't ask an LLM about its capabilities; by design they are unable to know what they're capable of. You can read this for more details:

https://arstechnica.com/ai/2025/08/why-its-a-mistake-to-ask-chatbots-about-their-mistakes/

OpenAI will try to coax it into saying things that are probably truthful, but ultimately an LLM doesn't know what truth is. They just spit out information that seems like the most statistically useful response.

On 11/1/2025 at 2:48 PM, BeyondTime said:

What I am interested in is about an earlier point of using an easy button to make something it probably took you years of hard work to learn how to do.

Actually something that's good to watch about this is Tim's Vermeer. There's a quote in it I like which is, "people think it's cheating because they think it's easy," when it's actually not easy. It's better to watch the movie to get the proper context, but the point was that this happens over and over in the history of art, such as how photography is easy compared to painting by hand. New technologies that come out often make replicating existing techniques look easy, but that's just the starting point. Photography inspired the impressionist art movement (like Monet's works, capturing light itself instead of getting fine details correct). That's something that photography didn't do. In the beginning, photography was just ha! we got a perfect black and white photo that's more accurate than any painter can make! But photography didn't stop there, it was just the beginning (I'm sure many here know how hard it is to produce good photos).

Generative AI graphics is mostly still in the "ha! easy!" phase. To me, using services like Midjourney or ChatGPT to make images is a dead end, it's the same as those early black and white photos. IMO the real power will be having an AI model running on your own computer that understands generic concepts like what a person looks like, what foreshortening is, perspective, etc, and then feeding it your own artwork so it replicates your art style. The result is that only you have a model that lets you control your artwork with words (either generating images from scratch or editing stuff you've drawn). That is EXTREMELY powerful, and is something that a talented artist can make great use of. This can already be done by using Stable Diffusion on your own computer and training it with your own images so it can replicate your art (it requires a LOT of technical and artistic knowledge, but it can definitely be done). Here's an understatement: Stable Diffusion is legally troubling.

We're going to fly past the "legally troubling" hump one way or another, we can't close pandora's box.

IMO Adobe's currently in the most legally clean position when it comes to generative AI tools since their model is based on their stock photography catalog. They own all the rights for everything they used to make it. Seeing the publishers/authors/artists fight the AI companies reminds me of how Polaroid and Kodak fought over instant photography patents. Kodak ended up losing, having to pay Polaroid almost $1 billion. Both companies became irrelevant less than 10 years later because of how fast digital photography advanced. Some of the people filing some of these anti-AI lawsuits have the best positions to develop beneficial AI, but I have a feeling we'll be stuck with trash.

There is literally only one AI thing that I appreciate existing.