[raincole]

As a (I dare to say) rather advanced hobbyist digital painter, I don't want generative AI. I want a good image search engine + database.

When I say "a woman looking outside from her seat on a train" I don't want it to generate such an image for me. I want it to literally give me existing images that more or less fit the description, with full credit. If it's from a movie I'd like to know the movie's name and timestamp and the actor's name and everything.

I know the difference between referencing and plagiarism. And if I'm going to cross the line, at least I know I am the one who makes this decision and takes the legal risk. With the current generative AI I don't even know who holds accountability. Likely nobody, or worse, an unaware me.

This is why I think the copyright law is very broken. If someone made the database I described above, their pants, house and first-born would be sued off. But I'd argue it's a not just more useful, but also more ethical product than generative AIs.

[TeMPOraL]

Both GenAI and that database could exist - they'd complement each other perfectly. You could use generative models in the privacy of your home, for reference or inspiration or to quickly explore possibilities, and then paint like you normally do - or even use generated images as they came out. When the time comes for you to publish your work, you could use the database with an image search engine, to compare your work against already published works, and determine if you haven't accidentally violated copyright, or created something that could be seen as plagiarism.

Similarly, other people could use that database to check newly published work, making it easier to detect and stop obvious copyright infringement.

Problem is, the copyright system - both as a body of laws and as a spirit and mindset behind them - prevents the database from ever being created. At least in any form other than "yes / maybe?" responses from comparing perceptual hashes; go beyond that, someone comes out of the woodwork, seeking royalties. Reverse image search engines exist, but are barely helpful because of that.

Anyway;

> And if I'm going to cross the line, at least I know I am the one who makes this decision and takes the legal risk. With the current generative AI I don't even know who holds accountability. Likely nobody, or worse, an unaware me.

Ultimately, you're still the one making the decision. No one forces you to publish whatever a generative model produced in response to your prompt. It's up to you what to do with the output. You also exercise creative control - both during and after generation.

The legal situation of GenAI in general is still uncertain - but at the very least, you're still in control of whether you're referencing or plagiarising in a moral sense.

[doright]

Coincidentally I was also looking for one of these, does this work?

https://lenso.ai/en/search-by-text?desc=a+woman+looking+outs...

[raincole]

Looks promising. I'll give it a try. Thanks for the suggestion!

[raincole]

The business model is... interesting, so to speak. It only asks me to pay when I reverse search for photos of real-life people. I guess it's that stalkers/people on dating sites are more willing to pay...?

[creer]

> I want a good image search engine + database.

You would think we'd be there already. But search as a field has died. Hopefully temporarily but pretty dead. This is interesting because good search is supposedly how Google make their profit. And for years now they have obsessed about anything but good search. It's now a given that a google search will give you anything BUT what you want. I know: there are upstarts - are the upstarts that good? For example, is there a great image search upstart?

[qbane]

My educated guess is that the profit for such a "magic" search engine is way too small than a magic "painter" that is more likely to replace a real artist. So the AI tech companies prefer the latter. It is all about the business model of the world we are in.

[TeMPOraL]

GenAI models are meant to create, not retrieve. The model being able to sometimes reproduce copyrighted material in training data is an unintended side effect, and usually requires the user to intentionally cause it to happen. Their business model is indeed a "magic painter" (and/or "magic writer", "magic coder", etc.), which is IMHO valuable and fair - again, it's meant to create, not reproduce or retrieve.

[qbane]

I indeed confused LLM/foundation models and GenAIs. They are both the current tech trend and both fall under the category of AI productivity tools, and I agree with your point.

[1vuio0pswjnm7]

"I want a good image search engine + database."

One could hire a software developer to write such a program. But, in general, software developers can be untrustworthy and prone to stealing ideas for their own selfish purposes.

Before Mark Zuckerberg dropped out of college, he answered a classified ad from someone in the real world^1 who wanted to create a website that could search a database of images. Nothing to do with "social media". IIRC the target market was the auto insurance industry (I could be wrong).

1. Not Winklevoss brothers. That is another story.

Zuckerberg never finished the job but he did create his own website which searched and served up images of students at Harvard. He was of course sued by this person after Facebook became popular.

[1vuio0pswjnm7]

Perhaps it is not a coincidence that so many software developers commenting online seem to have an extreme dislike for intellectual property.^1 It interferes with the stealing and copying they practice to compensate for their won lack of creativity. For example so-called "Big Tech" companies that do not produce any content despite billion of dollars at their disposal; instead they copy and store other peoples' work, so-called "user-generated content" (UGC).

1. With the exception of any intellectual property underlying software licenses. Microsoft wants a fair use exception for copyrighted works used by OpenAI but aat the same time it aggressively pursues copyright enforcement over its software, such as "Windows" and "Office".

[cwsx]

> One coulld hire a software developer to write such a program. But, in general, software developers can be untrustworthy and prone to stealing ideas for their own selfish purposes.

Ehhhh? Yes there are examples of that, as there are for any arbitrary group of humans you could select, but [anecdotally] I've noticed the opposite... it's not uncommon to find a passionate developer that's only interested in the challenge/problem solving aspect - it's a lot less common for say.. real estate agents.

I don't really get the point you're making beyond "people be greedy sometimes" (which I do agree with, don't get me wrong).