Progress on predictions means both getting better at prediction (learning) and applying that.

Also a decent definition of intelligence

I think for both contexts its far too simplistic to be more than a generalization and certainly for fun its a very local definition to serve Raph's ideas about what constitutes a game rather than encompassing enough to define it fully.

For intelligence for example you could have a PID controller where there is automatic tuning which would fit the definition of learning and application. But I don't think we'd call it intelligent outside of marketing copy.

A PID doesn’t get better at learning and applying predictions. I’d argue that to do that essentially indefinitely requires intelligence.

Hence mentioning a PID controller that has autotuning. Drop it in a new environment and it'll adjust. Drop it in another and it'll reconfigure itself.

That is not getting better at learning. That’s repeatedly re-learning in the same way.

Ahh, sorry we’re talking past one another then because I hadn’t twigged you were talking about getting better at learning because that’s not what I meant with my initial post! Although I can see why you took that from it.

I do like that meta observation though that not only do people get better at prediction through learning they can also get better at the rate at which they improve their predictions.

It’s careless of me to say it’s a definition of intelligence, but I do think that property of being able to improve how you learn and how quickly you learn (especially in response to adversaries doing the same thing) is a clear indicator of intelligence and there’s a good argument that that’s why we developed intelligence. These aren’t my ideas either, I’m just parroting what I recently read in the book “What is Intelligence” by Blaise Aguera y Arcas.

No, it's not a very local definition at all, it's actually a generalized definition for all forms of game and entertainment -- and art, even!

You seem to be assuming I have a reductive definition of game, when the definition given in the article is basically "anything people choose to play." See https://www.raphkoster.com/2013/04/16/playing-with-game/ which is linked in there.

I strongly disagree with lumping "intelligence" into the question though, so I am with you on that.

I can get better by getting more experienced without getting more intelligent.

True, but one definition of intelligence is the ability to deal with a novel situation. You can't get more experienced if you're "too stupid" to learn and adapt to the challenge.

Why do you think that accumulating experience and applying it to be better isn’t a mark of intelligence?

In simple, reductive terms:

Fun, among other things = remaining in the tight channel of flow, where your skills get challenged without ever reaching a point of frustration. Too little challenge = boredom.

Skills improve as they get challenged, i.e. when our prediction and pattern matching system receive enough feedback to improve upon our previous actions to get a more optimal outcome.

So, fun is (among other things) getting better at doing something, and as we get better, what was once a challenge turns easier, so a fun game needs to have a well-tuned difficulty progression to keep in pace with your improving skills.