WotC Would you buy WotC products produced or enhanced with AI?

[BookTenTiger]

Pretty sure no human would of made these.

People judged those things as "ai fails".


At it's simplest is

1. Humans give something a score.
2. Give AI a bunch of examples.
3. Spend a long time calculating how to get the highest score with that data. More compute allows for more data and nuance.
4. It can now rapidly and repeatedly do the thing that gives the highest score.

So if you want "imaginative" AI, then we just need to do is score a bunch of "imaginative" things. Though I expect a lot of disagreement on that, and there's not enough compute for individual tastes yet.

Note you can use negative scores too. So ranking stuff by how "slop" it is works too. (I.e. a smooth car ride is 10 points, a fender bender is -500 points, and a totaled car crash is -10,000 points).

I am 100% positive all those images can be, and in fact have been, imagined and created by humans before. You've got a variation of Mona Lisa (done a million times before), Wheelers from Return to Oz and a baked good in the form of an animal, which people have been making since baking began.

I just saw a character like your final two in the very imaginative, human-made web comic Kill 6 Billion Demons.

Without human creativity and innovation none of these objects could exist. I'm sure artists can use AI to add to their ideas, but without human creators AI can only feed on its own tail.

 

[soviet]

Pretty sure no human would of made these.

People judged those things as "ai fails".


At it's simplest is

1. Humans give something a score.
2. Give AI a bunch of examples.
3. Spend a long time calculating how to get the highest score with that data. More compute allows for more data and nuance.
4. It can now rapidly and repeatedly do the thing that gives the highest score.

So if you want "imaginative" AI, then we just need to do is score a bunch of "imaginative" things. Though I expect a lot of disagreement on that, and there's not enough compute for individual tastes yet.

Note you can use negative scores too. So ranking stuff by how "slop" it is works too. (I.e. a smooth car ride is 10 points, a fender bender is -500 points, and a totaled car crash is -10,000 points).

It was the best of times it was the blurst of times

 

[J.Quondam]

Pretty sure no human would of made these.

Really? Body horror has been a thing onscreen for decades, and far more imaginative that any of those pics. And people have been imaging weird hybrids since they started inventing mythology, albeit not crossing puppies and blueberry muffins (though I wouldn't be surprised if there's an anime out there that has done just that).

I mean, those specific images maybe weren't created by a human, but there's nothing about them that suggests a human couldn't have. Even the trippy DeepThought (its name, iirc?) image of the the Mona Lisa made of dog snouts is clearly procedurally generated. It's weird, but not particularly surprising to someone who's messed around with sophisticated math software in the last 40 years or so.

 

[Morrus]

So if you want "imaginative" AI, then we just need to do is score a bunch of "imaginative" things. Though I expect a lot of disagreement on that, and there's not enough compute for individual tastes yet.

I don’t think you understand what imagination is.

 

[mellored]

I don’t think you understand what imagination is.

Feel free to give me some examples then.

Preferably a millions of examples that I can train an AI with. Or even a few so I can learn.

 

[The Firebird]

You are also dehumanizing people who are trying to help other fellow artists, publishers, & hobbyists survive in increasingly hard times for many of us.

Perhaps you might want to consider having some empathy for those who are directly affected by recent developments, rather than throwing them under the bus in the name of ill-gotten "progress"...

I think it is possible to 1) have empathy for people whose careers will be damaged or destroyed by the introduction of new tools and 2) recognize that those tools can be used beneficially.

I see a lot of absolute statements, a lot of black and white lines in the discussion. I don't think that's useful for distinguishing between where AI is useful and where it's harmful.

I can't help but feel the criticisms about quality, like "AI will never do X", are wishful thinking. It would be nice if it couldn't, because then it wouldn't be a threat.

 

[Morrus]

Feel free to give me some examples then.

Preferably a millions of examples that I can train an AI with. Or even a few so I can learn.

Do you literally not understand what the word means? I thought you were doing a bit, but now I think you fundamentally don’t understand the word? Or you’re just doubling down on the bit. They do say that satire is harder these days because it’s so easily confused with reality.

Imagination isn’t like a colour. It’s the fundamental act of creation from nothing. It’s how new stuff comes into being. It’s the exact opposite of training something to reproduce it, by definition.

But you know that, right? You’re doing a bit? You’re trolling me?

 

[Art Waring]

What appeal to empathy can you make that those who lack empathy will understand? None.

People who lack empathy will not even begin to understand until they are personally affected by something. And even then it's only that they were personally affected because something about the lack of empathy also prevents generalizing one's own feelings to the feelings of others, much less acknowledging that others have feelings. So to a lot of absurdly pro-“AI” people it's all academic until they are the ones to lose their job, house, healthcare, access to food, etc.

They're this guy, basically.

I agree 100%, but my post was not an appeal to empathy (that was just one line of the text), it was an attempt to put the "Luddite" myth to rest. Too many people fling that word around as an insult, implying that those who are opposed to gen-ai are uneducated, knuckle-dragging evolutionary throwbacks that hate all forms of technology.

On the contrary, in my experience, the folks that are the most adamantly pro-ai are the ones who know the least about the true nature of generative-ai (just look at all those CEO's raving about AI, when they know absolutely nothing at all about what it is, or what it actually does, or even how to implement it in their business). As soon as you become aware of this, well you have a choice...

You can choose to be aware of the situation, and try to help improve things for people affected by it,

Or, you can choose to put your head in the sand, throwing other people under the bus for the sake of your own personal monetary gain. (as many of those who are pro-ai can't help themselves, continually pointing out the fact that its "free" without looking at the incredible amount of resources it takes to keep gen-ai running).

That choice has already been made in many cases, changing minds is hard.

Some, however, are unaware that they have made a choice at all, as they are repeating, like a parrot, talking points that have already been covered elsewhere, and end up sounding a bit like a broken record. (the "AI learns just like an artist does" argument is so tired at this point, as applying human behavior to machines is rife with problems, and it is factually inaccurate to compare how computers process data with how humans create works of art).

 

[Faolyn]

I legitimately wonder what the"never AI" folks plan to do with the massive AI integration into every piece of productivity software they are likely to use, without the ability to opt out.

I turn it off where I can and don't use it where I can't.

 

[ezo]

What human artists do is fundamentally different than what current gen AIs do.

For now... And yeah, it is just scale. You brought it up, after all.