Email or username:

Password:

Forgot your password?
13 comments
Matt Γ— πŸ‡³πŸ‡΄ πŸ‡ΊπŸ‡¦

What’s interesting to me here is that both the big name pop stars, Lizzo and Harry Styles, are actually less detailed illustration styles and less accurate representations than the images of serial killer Harold Shipman and internet viral video sensation Ronnie Pickering.

Alex@rtnVFRmedia Suffolk UK

@matt this doesn't surprise me one bit - if the AI is using a corpus of photos from news websites in the UK there are going to be far more of criminals (Pickering has been arrested and put before Court at least once in recent years) than of pop stars, as this is the greater balance of content in our news (fuelled by the algorithms carefully monitoring what gets "engagement").

Matt Γ— πŸ‡³πŸ‡΄ πŸ‡ΊπŸ‡¦

@vfrmedia Umm, but the internet has a million more pictures of Harry Styles than Harold Shipman. But I guess that’s also the point; too much selection confuses it slightly? Maybe.

Alex@rtnVFRmedia Suffolk UK

@matt but is the AI using the "whole internet", or a limited/curated source of photos including UK news sources? (presumably it has to comply with some copyright agreements?)

There's also a whole load of "true crime" blogs that repeat all the major cases from UK with photos, we don't have as strict privacy laws as other Northern European nations and crime reporting is very popular here (if asked, I could probably name and recognise more criminals than modern pop stars..)

Matt Γ— πŸ‡³πŸ‡΄ πŸ‡ΊπŸ‡¦

@vfrmedia I wonder if there’s even restrictions legally on what the AI can learn from… Copyright wouldn’t cover it, surely? I mean it’s not like the bot recreates an image - just might use it as a reference before it spits out any pixels.

There’s so much I don’t know about this stuff. πŸ₯²

pettter replied to Matt Γ— πŸ‡³πŸ‡΄ πŸ‡ΊπŸ‡¦

@matt Derivative works are covered by copyright, and it's hard to be more derivative than thoughtlessly mashing together the average of prior works. @vfrmedia

Go Up