Mateusz Bagiński

Posts

Sorted by New

Wiki Contributions

Comments

Science in a High-Dimensional World

The great challenge is to figure out which variables are directly relevant - i.e. which variables mediate the influence of everything else.

Is this equivalent to identifying the Markov blanket of the phenomenon being studied?

Salvage Epistemology

I'm not sure how the right decision process on whether to do salvage epistemology on any given subject should look like. Also, if you see or suspect that this woo-ish thingy X "is a mix of figurative stuff and dumb stuff" but decide that it's not worth salvaging because of infohazard, how do you communicate it? "There's 10% probability that the ancient master Changacthulhuthustra discovered something instrumentally useful about human condition but reading his philosophy may mess you up so you shouldn't." How many novices do you expect to follow a general consensus on that? My hunch is that if one is likely to fall into the crazy, they are also unlikely to let their outside view override the inside view, assert "I calculated the expected value and it's positive" and rush into it. Also2, how does one know whether they are "experienced enough" to try salvaging anything for themselves? Also3, I don't think protecting new rationalists in this way would be helpful for their development.

To reduce the risks pointed out by OP, I would rather aim at being more explicit when we're using salvage epistemology (here just having this label can be helpful) and poke around their belief system more when they start displaying tentative signs of going crazy.

Predicting a global catastrophe: the Ukrainian model

You're probably right. I was myopically looking only at the rightmost portion where GJP updated to ~99% a bit quicker. It also seems like GJP had a more erratic trajectory than Metaculus.

Predicting a global catastrophe: the Ukrainian model

If I remember correctly, GJP superforcasters were similarly successful, although they were a bit slower.

Actually, GJP forecasters updated a bit more quickly than Metaculus [EDIT: probably not, look up the reply below]

Let's buy out Cyc, for use in AGI interpretability systems?

My superficial understanding is that Cyc has two crucial advantages over all current knowledge bases / knowledge graphs:

  1. It is much, much bigger
  2. Predicates can be of any arity (properties of one entity, relations between two entities, more complex, structured relationships between N entities for any N), whereas knowledge graphs can only represent binary relationships R(X,Y), like "X loves Y".

If I understand it correctly, then Cyc's knowledge base is a knowledge hypergraph. Maybe it doesn't eventually matter and you can squeeze any knowledge encoded into Cyc's KB into ordinary knowledge graphs without creating some edge-spaghetti hell.

Moloch and the sandpile catastrophe

Typo: the Seven Years' war ended in 1763, not 1753.

Antimemes

Different cultures have different antimemes. The more different two cultures are from each other the less their antimemes overlap. You can sweep up a mountain of antimemes just by reading a Chinese or Arabic history of civilization and comparing it to Western world history. You can snag a different set by learning what it was like to live in a hunter-gatherer or pastoralist society.

Anti-memes as shibboleths? If I see that you share the same weird belief/cultural practice/whatever that provokes such a strong self-suppressing response that nobody would acquire it unless they (a) are already immersed in that anti-meme's culture or (b) make some substantial deliberate effort to become a part of that culture.

March 2022 Welcome & Open Thread

I'm subscribed to So8res's posts and over the last ~2 days have been getting messages that "So8res has created a new post: X" where X was a 7-year old post.

Russia has Invaded Ukraine

You might want to edit this for clarity: in English Estland is Estonia and Letland is Latva. This was not immediately obvious to me on first glance.

Load More