Hacker News Reader

Comment by dwohnitmok | Hacker News Reader

> ...while higher-level artistic work will be done by artists, assisted by AI. Artists can get inspiration from AI artwork and reduce some tedious parts of their workflow.

I'm having a hard time seeing how increasingly better and freely available AI artwork doesn't just completely replace artists (and more generally how increasing AI capabilities don't just replace any career that does not involve some degree of skilled manual labor, at least until AIs get robust manual capabilities). In particular "higher-level artistic work" would seem to be that work which rises beyond what most artists do today. Prompt engineering is a far cry from making a painting oneself and there are likely many people who enjoy the latter who do not enjoy the former.

View on HN

> I'm having a hard time seeing how increasingly better and freely available AI artwork doesn't just completely replace artists

I agree that AI engines like DALL-E seem like a threat to many commercial artists, especially those whose name and personality are not closely associated with their works. Magazine illustrators are one example.

For some artists, though, the artist’s status as a human being with a unique identity is a big part of their appeal. That’s one reason why there’s such a huge price difference between, say, a painting believed to be by Van Gogh and a high-quality painting in Van Gogh’s style that is believed to be a forgery.

My daughter is a successful commercial artist, and I have been following the AI advances partly out of concern for her future. As it happens, she has been able to associate her name and personal identity even with her commercial works, and she is skilled at working closely with her clients to produce the kinds of illustrations they want. For the time being, at least, her business seems safe. I wouldn’t want to make a prediction about the long term, though.

> she has been able to associate her name and personal identity even with her commercial works

Doesn't this just mean that her future former customers will add "in the style of [her name]" to their prompts?

Possibly in the future, but not yet.

I just tried several DALL-E prompts with “in the style of [my daughter’s name],” but the results looked nothing like her work. Maybe OpenAI hasn’t crawled her website [1] yet.

Also, her commercial work usually includes lettering in English (and sometimes in Japanese), and DALL-E is not able to produce words or sentences yet.

[1] https://www.saragally.com/

> Also, her commercial work usually includes lettering in English (and sometimes in Japanese), and DALL-E is not able to produce words or sentences yet.

This is purely an issue of scaling and is no longer true. Other text-to-image AI have already gotten around that and produce accurate text. See e.g. https://imagen.research.google/

I can imagine a future that the artist could "license" Dalle-2 to generate art in their style and they actually get some commission for it. Like the AI can work as the artist's assistant to produce art at scale while she gets benefits too if it makes sense.

If I judge by the examples posted on OpenArt, Dall-e is currently perfectly happy to produce art "by [artist name]". What would they gain by starting to pay commissions? I doubt that will happen unless some legal action forces them to do it.

Even in that case, how would you decide which artists get the commission and which do not? They can't possibly track down authors of millions of artworks across the globe they scrapped when training the model. It seems to me that paying commissions to anyone would just mean that they acknowledge the questionable copyright status of the artworks produced by Dall-e.

Reply on news.ycombinator.com