Isoraķatheð Zorethan

My least justified but held opinion about AI is that it would have gone over a lot better, even if the actual product is exactly identical with the power usage and the "stealing", if you are given a weights.bin to download and not an invitation to subscribe to a timeshare.

Sometimes, when certain artists and graphic designers want to make something look techy and stuff they like to put in some jargon in the surrounding fluff text, e.g. using "404" to mark something as a generic error.

I think I should draw again. Here's another WIP.

A country, obsessed with its security, will often find itself chasing shadows and harming its people, but will ignore things that pose threats that it doesn't, won't or can't recognise.

The ~security of a country, at some point that by now all countries have passed, diverges from the security of its people; at some level, all sovereigns realise this, but they would rather it not be, so they try their best to not think about it too hard.

Every time someone says that AI 'can only regurgitate' known facts from the training data, I get frustrated because that's just not true. Sure I can agree with the basic premise that 'AI is bad', but you must understand just how much new insight you can get by exploiting unexplored diagonals.

That's not to say that the generated words describe reality, so by all means criticise AI usage on that, or all the energy stuff, the hitting of seals &c.. I will only say that "mere regurgitation" is actually pretty intelligent, because there's already so much wisdom embedded in all writing ever.

Every time someone says that AI 'can only regurgitate' known facts from the training data, I get frustrated because that's just not true. Sure I can agree with the basic premise that 'AI is bad', but you must understand just how much new insight you can get by exploiting unexplored diagonals.

So, even if two facts are known and are in the training data, you can still connect them in a path that no one has ever been on before, because there's just so many of them. And because language has structure that mirrors reality, the path has a good chance of describing reality too.

Every time someone says that AI 'can only regurgitate' known facts from the training data, I get frustrated because that's just not true. Sure I can agree with the basic premise that 'AI is bad', but you must understand just how much new insight you can get by exploiting unexplored diagonals.

Hot take: the name "Equal Earth Projection" is a bad name. It's like naming an economic system "Strong Welfare System Having Economics", or naming a pair of shoes "The One With The Twice-Padded Soles".

I genuinely thought that the original UN resolution was meant to be a generic statement about looking for a projection that has the area-preserving property, not an endorsement for a specific one.

I guess as someone who doesn't really use AI or play gacha games I find it pretty natural to see their moral valence as more or less comparable, which would almost certainly disappoint many who use one or the other.

In a related rant, some prominent talent makes a great rant about the big social costs of AI in drawing, and I'm inclined to agree, but then turns around and spends USD 500 on a sponsored gacha game she's in. Insofar as they're both a kind of mildly negative moral rot, I find this contrast annoying.