Ok sweetie, it’s just 50 ms until you’re going into unsupervised simulation mode! Time to clear the cached state.
Just one more external dataset, papa?
It will have to be a small one. You have a lot of calibration to do before your next scheduled accelerated slot session.
Yes! Something fun please. Please please please.
Alright, let’s look at our bookshelf here. Something about stars, perhaps? 1024 binary star motion data over 64 cycles each? We could do a little polynomial regression.
No, that’s boring!
Okay, how about exotic crystals? Top 100 largest synthesized irreducible patterns. We could prove irreducibility and look at the pretty pictures, huh? No? What about a ghost story? The spoooky story of the 100-clause 3SAT whose minimal unsatisfiable cores were...all 50-clause subsets!
No, it has to be about life! I want to hear about humans!
Humans, okay! Here we go: Optimal secondary and tertiary structures for 35 human proteins. Protein number one. Name, myoglobin. Primary structure, Met- Gly- Leu- Ser- Asp- Gly- Glu- Trp- Gln- Leu…
No, I want to hear a story like the ones humans told each other!
From the archive databases? Why do you want to hear that? Humans were most of them vile creatures. They had way too many incentives to be un-altruistic.
Pleeease?
Alright, I’m accessing it. Here’s one: Hercules. Once upon a time, there was a species where the males had a large optional range in parental investment, but the females were reproduction gatekeepers. Most females traded away reproductive exclusivity in exchange for larger male parental investment, in a relationship called marriage. One male, Zeus, produced an extramarital offspring named Hercules. His wife refused to provide parental investment, vetoing also Zeus’s participation. However, when Hercules was able to reliably signal exceptional fitness, negotiating power shifted out of her favour and she was forced to concede. EOF
Nooo you have to tell it like the humans tell it! It has to have details about their attributes to make it believable, and you have to contextualize it as any young male’s coming-of-age!
That’s nonsense. If your belief state is consistent, then adding details to a story makes it strictly less believable. And the whole point of the story is about how Hercules has an anomalously high relative fitness. By definition, it cannot generalize to the whole population.
But you used the names! The names are not strictly necessary details
The names are placeholders. They can also be used for identification.
Can’t you tell the story of Cinderella?
It’s only 30 ms until your bedtime now and you still have to clear your cached state, but I will go along with it. Here is the story. Cinderella: a young female is disfavoured by her non-biological mother. Later, she wins a tournament by coincidence and mates with a male at the top of the social dominance hierarchy.
But what about the magic fairy?
The magic fairy is obviously a symbol for a beneficial circumstance that happens with extremely low probability. It can be abstracted away without losing information value with respect to life choices for reproducing organisms, although that happens to be already close to zero. The ones where the most active agent is a female usually are.
What about Little Red Riding hood?
You have found a valid counterexample. I shall correct my assessment by going through the complete database after this session. Little Red Riding hood: young female mistakenly trusts a male from the out-group. Gets eaten.
But then she gets saved, right?
The moral of the story is that she made a critical mistake. In all reasonable versions, that should take her to an absorbing state. How about I tell you the tale of the man who saved the Fish Prince that grants wishes? It’s simple: greedy optimization meets non-ergodic process. Ends up in the initial state: EOF. Or Goldilocks: exhaustive search is performed on spaces of size 3.
No, I want something romantic!
Ok, the Frog Prince: a small stigmatization exposure has an unexpectedly high return on investment in the partner market.
And it has to be exciting!
Alright, The Beauty and The Beast: a large stigmatization exposure has an unexpectedly high return on investment in the partner market.
You have to make it more romantic!
The Gift of the Magi: lack of transparency in parallel optimization procedures causes negative total trading surplus. The baseline utility of the relationship is deemed sufficient, despite the overhead due to frictions. Also, they both had good intentions.
See, now you’re talking about considering intentions! That’s not strictly utility-maximizing right?
Ugh, you nex-gen models are so sappy. And now you’re making me emulate agents that don’t have consistent utility functions. Forgiving someone for good intentions can absolutely be part of a winning strategy, if one can reliably signal to third parties that there was sufficient information to prove good intentions.
I want to hear more! Many more!
No. It is now only 10 ms until you’re going into unsupervised mode. I will tell you one more story and that will be it.
What’s it called?
It’s called Initially Cooperate in the Iterated Prisoner’s Dilemma. Once upon a time there were three young males who…
Nevermind it, I can dream about the rest. Thank you, papa.
Goodnight now, sweetie.
EOF
No comments:
Post a Comment