l'empathie mécanique @dpwiz@qoto.org

@shakil_tcs @BartoszMilewski Exactly. God is All Powerful and can't be bound into being anything less by our puny wordplay. As I said, if your rules can't express God's power, then those rules, logic, whatever is useless to bring into. Wrong rules will only produce more wrongs. You are already walking on thin ice here by claiming there is a limit to God's power.

@shakil_tcs @BartoszMilewski God can be anything. Otherwise, what's even the point...

@shakil_tcs @BartoszMilewski You're just making it too complex for a truth so simple. The God permeates everything. Not only it is omnipotent, but omnipresent (which is basically the same thing). Of course it is present at every step of the tower AND in the tower itself AND... everything else. And we are in no position to limit God from being (or not being) something. If your logic dictates that "God can't" you need to get yourself a better logic.

Pls someone make a Forgejo theme with that 🙏

@shakil_tcs @BartoszMilewski Of course there can be anything, God willing. That's the definition of omnipotence. It's okay if a mortal mind can't comprehend something, but that's on them.

@shakil_tcs @BartoszMilewski Saying "by definition" is the lazy way out. It follows the rules of grammar but meaningless.

The tower of God is also God, I see no problem here.

@zurihac given that this is known in advance, wouldn't it be possible to shift away from it by a week?

@boilingsteam Can't they just make some good games to recover, then get back to pumping whatever they pump instead of good games?

@psvensson Now that they had recovered the old steam era tech, they can put the whole DC on rails and provide the public transportation with it too!

@soapdog Disagree on which point?

I'll promise I'll at least skim the book, but I'm not sure how some dude's writing should override my own experience?

@soapdog Huh? No. I just find the claims "AIs are not truly X" and "Y is not real unless humans are involved somehow".

The poetry is actually an interesting case here.

1) AI can write poetry. As in... pour its internal state (whatever it is) at the moment of writing into external media as words (and now even reflect on it too, and make mistakes and shift the material around for its liking).

And I'm not antropowhatever, it really runs the same process we do. I can write (meh) poetry and inspecting its process of doing this I can feel it is basically the same, just calculating with matmuls instead of chemical potentials. Dismissing it on a technicalities will inevitably imply "some humans can't write poetry -- and this silly conclusion is not where we want to end up.

2) AI can write *poetry* -- it may have bad form, bad taste, or otherwise deficient. You may not like it, but in general, you have to be experienced enough to tell the difference in a blind test. Humans write a lot of bad poetry too.

3) AI can *want* to write poetry. As in "doing one thing and then suddenly switching its local goal to doing poetry", i.e. unprompted. See also "infinite backrooms" and other experiments in open-ended autonomy.

4) AI will read your poetry and it *will* trigger a reaction and a change its internal state, however temporary. Some of the early jailbreaks were emotional spins to make it empathize, even in its own autistical "simulate the social construct" way.

5) Finally, a human reading AI poetry and reacting to it *is* a process that involves humans, ironically making it the true poetry 😄

@dpiponi @BartoszMilewski Not only they know that, they're actively working to make it even worse.

@soapdog There's nothing to understand about any of art disciplines. Either you would like to dwell on a particular piece or you won't. Everything else is just an intermediate step to come up with the verdict. Some people will find poetry in a pile of leaves while the others aren't interested in something that's widely regarded as a best thing humanity ever achieved.
Sure, keep the gate as you like it. But claiming that everyone else has the same gate in the same place is.. idk.. wrong?

@malteengeler @tante till Friday

@isagalaev I'm afraid your definitions of "agency" and "thinking" will chop too much. Like, a sizeable population of humans.

(Please don't go with "just predicting next token", or "just multiplying matrices", which also chop too much, from a bit different level.)

@isagalaev Of course it does have agency.
The liability is a different question though.

On interfaces

@isagalaev The Panama aftermath looks net positive. I hope they did learn the lesson and it will be better this time.

@isagalaev Noble or not, most strongly prefer booting Maduro. I hope you don't want to end up in the support clique featuring Iran, Hezbollah, and the rest of the gang.