5. Willing to look foolish in socially subversive ways = can learn exploratory new things, with weird outcomes that change valuation of rewards/penalties for all, while potentially catalyzing weird schisms and fractures in the social landscapes. Inviting outcaste status.
Conversation
People without experience in explicitly codified caste systems don’t realize that the biggest punishment is not being lower than another caste. It’s being left out of the caste calculus entirely, hence ‘outcaste.’ You can only choose to subvert the system, not compete within it.
2
4
61
But so long as the caste crowd doesn’t try to stop you beyond just excluding you from their markets of social/political/economic rewards, it’s actually the least constrained learning mode. You’ve got no “face” to lose. You’re unseen anyway.
2
4
45
A good deal of the motive for systematic structural oppression is in fact fear of the great freedom that comes from being cast out of ritual and institutional constraint frameworks on learning. You can just be kinda wild and unschooled and antifragile about it.
1
5
36
The culture war (ie internet of beefs) is almost entirely levels 3 and 4. Very little by way of breakout level 5 dynamics that ignore and subvert social lines of conflict.
1
28
Among people young enough to be tech literate (under 60) being a lurker, or anon, or restricting social activity to media like product/restaurant reviews is often a sign of unwillingness to go past level 1/2. Waldenponding is typically a 1/2 behavior since meatspace is mostly 1/2
1
25
This was actually inspired by trying to come up with a scale for machine learning, generalized from stuff like 5 levels of autonomy for driverless cars. Most AIs learn from human-generated or tagged data, so they are level 3 by default and can't break that ceiling.
1
22
AlphaGoZero hits level 5 in a very constrained environment. Most older chess/Go type programs were level 4.
Because "foolishness" is encoded in data, you don't need the learner to have human conceits and vanities.
2
1
16
And this is not a strict hierarchy of enlightened learning or something. As a 1-d model it is not good for that. 2 other factors: age of domain and existential risk (as opposed to social risk) factor into things.
1
11
In very old and/or very dangerous fields, where sheer size of training data corpus from history and/or mortal risks far outweighing risks of looking foolish, a level 1 or 2 learner can be far wiser than someone going for level 5. Fools rush in where angels fear to tread etc.
2
22
A great example is the Perseverance Mars rover. It uses an ancient processor that couldn't even run GOFAI learning algorithms, but is radiation hardened. It does no "learning" and many production cars on earth do far more. It's almost entirely level 1 "unwilling to look foolish"
Replying to
But on Mars, you don't have the luxury of a lot of trial and error and do-overs. So the entire weight of everything learned in 70 years of space flight and several hundred years of engineering goes into building a 1-shot high-risk system. An enormously capable L1/2 system.
1
17
Perseverance is an "outcaste" in a car caste system defined by the scale from Kias to Teslas, BMWs and Benzes, to super-luxury italian cars etc 🤣
But it costs a gazillion times more, and lives on another planet where the money that cost is priced in is entirely worthless
2
12
Basically caste systems also need to be near monopolies on critical resources in order for outcaste to be a costly condition. To the extent they are not monopolies, they face risk from outcaste subversion. "You must play, we control who can play." twitter.com/thiirid/status
This Tweet is unavailable.
1
1
9
This thread also got me thinking differently about heresy and the idea of being a contrarian or heretic. It's trying to have outcaste freedoms while enjoying incaste access too. Like Michael Scott saying, "Ryan has book smarts. I have street smarts... and book smarts."
1
1
12
"Heretic" is basically an intrapreneur pretending to be entrepreneur at societal scale. Real heretics live in slums or in cabins in woods smelting their own ore. Human equivalents of Mars.
2
14
