Live data from Hacker News

AI doppelgänger experiment – Part 1: The training

julienposture.substack.com

131–137 of 137 posts

Re: AI doppelgänger experiment – Part 1: The training

#131

Earlier quoted context omitted.

Ok let's try a different argument. You think the world would be a better place if Monet was able to sue Renoir out of the impressionist movement? Or if Spielberg was able to sue director's trying to emulate the Spielberg oner?

Impressionism has scope. Most people can tell a [famous] Renoir from a Monet and would not think Renoir copied Monet (in general). The devil is in the detail, courts have too decide what is too close and what isn't, but it's very much 'I know it when I see it', I feel. Allowances to be made for distinctiveness. I'm not a movie buff, I know a few films from Spielberg's oevre, but I've never seen anything that I though…

I'd like to stick with Impressionism. Perhaps you think Renoir is distinct, and sure, he is. Now what about the hundreds of non-famous artists that are not Renoir and are instead just trying to make novel art in exactly the style of Monet? You think their original works are illegal? You think they are more illegal than an ai generated monet painting that uses a different color palette to be distinct?

I'm not very interested in debating this further because I think we just disagree. I'm glad this is not the way copyright law actually works.

Re: AI doppelgänger experiment – Part 1: The training

#132
post #58
post #56

I only have a Ryzen 5 3600x and a 7650xt, and downloaded AMD's LLM tool. It runs, and chats. I have thought about training it on the 4500 pages of LJ entries I saved to PDF (I know I'd have to extract the text, and then spell check that puppy. or not. Maybe I should just leave my misspells if it's a doppleganger). But it seems it'd take a few weeks of running nonstop to train? Everywhere I'm reading says it's hard to…

If you hate AI, why do this? If you just want to do it, why not just do it?

I should clarify - I'm ambivalent on a lot of the current structures (political, economic) that this is being created in. It's less about "AI" and more about how it's being shoved down our throats. I have ethics concerns (even though I understand and am slightly sympathetic to the arguments that "training" an AI on data is like a human studying previous works - though I don't think that's what's really happening). I have concerns about this being done "at scale" (especially seeing what Nvidia is doing to build giant massive data centers - though - as with most tech "it gets cheaper/cleaner/more efficient" so I can understand over time this could go down).

Being local reduces some of these issues... But even then I'm not sure I trust the big actors the renege on the original deals and start hoovering what they can.

That's why I'm willing to play with this locally on my own end just for funs while not necessarily buying into the hype, or trusting the players/creators that they have the best intent besides more exploitation of natural, human, and data resources.

Sometimes it's ok to have complex opinions, and admittedly I was a bit strong in my condemnation above, but also brevity in comments is a virtue, better to underexplain at first than overexplain and clarify if needed. (He says as he spends multiple paragraphs over-explaining in a followup)

Re: AI doppelgänger experiment – Part 1: The training

#133

Earlier quoted context omitted.

Not copy the work. Copy the style. Because style is not copyrightable. What I'm saying is that if you believe that your personal style is your valuable asset and choose to protect it by saying people can't train on your images; you ought to be aware that it's trivially easy to build a model that replicates your style and does not require your images.

Hiring someone for work is not trivial, though it is practical.

It's trivial compared to hiring a specific person

Re: AI doppelgänger experiment – Part 1: The training

#134
post #48

Please is there anything like this for writing? I am serious, I do write and want to experiment. Please would you have any suggestions for me? Thanks for your time.

Writing is the simplest of all. Really trivial actually. I create new styles almost every other day. See the latest entry on my blog for a style i created last week, and this video [1]. Different styles can be created by taking advantage of the context window. 1) Take some random style of a writer, a blogger, or even a website. 2) Generate some examples you may like, and annotate them, something like: Example Chapter…

Very impressive poetry of the gold markets in your blog! Wow. I am a bit speechless about the potential of all these.

Re: AI doppelgänger experiment – Part 1: The training

#135
post #48

Earlier quoted context omitted.

Writing is the simplest of all. Really trivial actually. I create new styles almost every other day. See the latest entry on my blog for a style i created last week, and this video [1]. Different styles can be created by taking advantage of the context window. 1) Take some random style of a writer, a blogger, or even a website. 2) Generate some examples you may like, and annotate them, something like: Example Chapter…

Very impressive poetry of the gold markets in your blog! Wow. I am a bit speechless about the potential of all these.

Infinite potential for sure. You can instruct it to be more terse, poetic, avoid pronouns when possible, dial up the humor and so on. Take also a look at David Shapiro's use of Claude and GPT for writing [1].

My process however is different than anyone else. By putting the machine to describe the writing, i use the keywords, remove one i may not like, and add one keyword of my choosing. Then i generate some more Example Paragraphs or Example Chapters and do the same again.

I slowly push the machine to generate examples until i am satisfied with the result. Then i go on and add my characters, and at last use the last bunch of examples to generate my story. It is not easy at first, but once i got the hang of it, now it is trivial.

[1] https://www.youtube.com/watch?v=bkbDgNjyDTM

Re: AI doppelgänger experiment – Part 1: The training

#136

Earlier quoted context omitted.

I swear there's a personality type that simply can't resist injecting their politics into unrelated conversations. It's truly fascinating.

I don't think that the opinion "70s intellectuals' love of the genocidal Khmer Rouge is bad" is "my" politics. I think this is a commonly held opinion actually.

https://www.jstor.org/stable/25006809

https://www.counterpunch.org/2021/09/10/noam-chomsky-and-the...

stop spreading lies that have been debunked for decades, not everyone share your inability to parse information from outside your ideological bubble

Re: AI doppelgänger experiment – Part 1: The training

#137

Earlier quoted context omitted.

Impressionism has scope. Most people can tell a [famous] Renoir from a Monet and would not think Renoir copied Monet (in general). The devil is in the detail, courts have too decide what is too close and what isn't, but it's very much 'I know it when I see it', I feel. Allowances to be made for distinctiveness. I'm not a movie buff, I know a few films from Spielberg's oevre, but I've never seen anything that I though…

I'd like to stick with Impressionism. Perhaps you think Renoir is distinct, and sure, he is. Now what about the hundreds of non-famous artists that are not Renoir and are instead just trying to make novel art in exactly the style of Monet? You think their original works are illegal? You think they are more illegal than an ai generated monet painting that uses a different color palette to be distinct? I'm not very int…

"Novel art" in exactly the style of Monet, is not entirely novel. I think it would be reasonable to allow Monet to have sole right to produce his style for 7 years or so, and then it enter the public domain. That rewards coming up with a new artistic style, which I feel is at least as valuable to reward as coming up with a new piece of artwork in an established style.

Again, my art history is weak, but when Picasso came up with his first cubist work, the cubism part of that work seems as valuable, and worthy of reward, as the actual instance of cubism that the work embodies?

There's a generally derivative style that follows the spirit of the age, and then there's a clearly copied style from a specific maker/creator/artist. It seems poor to protect the individual instance of an inspiring new style, but then not reward the creator of that style, not even with an acknowledgement.

FWIW, I'm not sold on either side of the argument. To me that's not how online discussion works. Also, I absolutely would not want copyright terms of the length we have to protect anything like this (nor anything really, they're obscenely long).

I would not of course want to inhibit the natural progression of styles, Monet's landscapes bore some similarity to JMW Turner, and a Chinese style called Lingnan, but this is a different thing to taking Monet's pallette used for water lilies, then painting a similar subject in the exact same style. If you do that immediately after Monet first presented his style for water lilies then I'd at minimum want you to say 'in the style of Monet' as a subtitle, and preferably have to pay him a little of your profit for the first few years.

Thanks for your interesting, thoughtful responses.

Post reply on HN