Summa Bayesiana

Two passages that say that you already know the truth (once it has been revealed to you), but you have to search for it anyway:

Thomas Aquinas, On Boethius on the Trinity, part 1, question 2, article 3, translated by Rose E. Brennan: "If, however, anything is found in the teachings of the philosophers contrary to faith, this error does not properly belong to philosophy, but is due to an abuse of philosophy owing to the insufficiency of reason. Therefore also it is possible from the principles of philosophy to refute an error of this kind, either by showing it to be altogether impossible, or not to be necessary."

Eliezer Yudkowski, Rationality, chapter 188:

"But perhaps it is not quite as exciting to see something that doesn’t look Bayesian on the surface, revealed as Bayes wearing a clever disguise, if: (a) you don’t unravel the mystery yourself, but read about someone else doing it (Newton had more fun than most students taking calculus), and (b) you don’t realize that searching for the hidden Bayes-structure is this huge, difficult, omnipresent quest, like searching for the Holy Grail.

It’s a different quest for each facet of cognition, but the Grail always turns out to be the same. It has to be the right Grail, though—and the entire Grail, without any parts missing—and so each time you have to go on the quest looking for a full answer whatever form it may take, rather than trying to artificially construct vaguely hand-waving Grailish arguments. Then you always find the same Holy Grail at the end."

Feyerabend and OpenAI

What we are seeing in the artificial intelligence field, specifically with OpenAI researches, resonates with Feyerabend's description of how science proceeds in Against Method. Here are some quotes from Feyerabend, and how they apply to AI in 2021:



  • "theoretical anarchism is [...] more likely to encourage progress than its law-and-order alternatives." : some AI researchers criticize OpenAI because they do not follow the "right" method (first you develop an original theory, then you test it), but rather see what happens when you give a lot of compute to an old theory; this "wrong" method is yielding the best results ever (progress).

  • "There is no idea, however ancient and absurd, that is not capable of improving our knowledge." : neural nets were considered obsolete a decade ago, now they produce the best (state of the art) results.

  • "To start with, there is the problem of telescopic vision. This problem is different for celestial and terrestrial objects; and it was also thought to be different in the two cases. It was thought to be different because of the contemporary idea that celestial objects and terrestrial objects are formed from different materials and obey different laws." : curiosly, the difference between what happens on the Earth and celestial objects is used as an image of what the OpenAI's approach does not get: "[...] we find this warning that our preferred approach to AI, in our case deep learning, may be like: '... trying to get to the moon by climbing a tree; one can report steady progress, all the way to the top of the tree.'" (And, specifically on GPT-3, here.)

  • "Galileo prevails because of his style and his clever techniques of persuasion, because he writes in Italian rather than in Latin, and because he appeals to people who are temperamentally opposed to the old ideas and the standards of learning connected with them.": OpenAI prevail (?) because of their style and their clever techniques of persuasion (examples in blog posts), because they write blog posts and not just papers, and because they appeal to non-researchers.

Go as a metaphor - Using what has been done

Why in go, even though everyone has the same means, are results different? Because a player places his stone in an intersection such that it, combined with the stones already placed, creates a better pattern, i.e. such that it makes a better use of the stones that have already been placed. Also, as soon as it is placed, the stone becomes an already-placed stone, so it is part of the existing pattern that can be used in the next rounds.

In other words, in go your situation gets better if you use what has already been done, repeatedly.

The same is true generally about human existence according to Karl Popper, who in The Self, Rationality, and Freedom (in Knowledge and the Body-Mind Problem, page 140) wrote: "You can do your work, and, thereby, grow through your work so as to do better work - and grow again through that better work, and so on. [...] The incredible thing about life, evolution, and mental growth is just this method of give and take, this interaction between our actions and their results by which we constantly transcend ourselves, our talents, and our gifts."

Decima azione

D: Chi e' il nano che fa il pittore ma non e' particolarmente bravo? Insomma, quando vedi i suoi quadri, non ti fanno ne' caldo ne' freddo.

R: ...Tiepolo.

13 = Evil

A TV series about a short-haired female psychologist and a Catholic man who is about to take an important step in the clergy hierarchy. They investigate seemingly supernatural events that may have a psychological explanation but are eventually part of a world domination plan. Meanwhile, they develop romantic feelings for each other... Am I talking about Evil or Il tredicesimo apostolo?

Nononano

Quando il filone della miniera si esauririrà, per avere diamanti da vendere i nani dovranno procurarseli in modi meno legali. Ci sarà bisogno di un nano ancora più basso degli altri, che riesca a passare per qualsiasi portello e cunicolo fino ad arrivare al caveau con i preziosi. Insomma, ci sara' bisogno di Botolo.

Go as a metaphor - Sacrifice is necessary

Even though I have written about similar principles ("you cannot have it all", "getting Y by skillfully doing non-Y", "the same thing can be a means and a goal, but not at the same time"), I think this is a bit different. In go, under the Japanese rules, you score one point for any empty intersection you surround with your stones. However, these stones also occupy intersections, which, as a consequence, can't score you any point.

In other words, in go sacrifice is necessary.

The same principles also applies to lossless compression, as any algorithm that makes some inputs smaller (so to speak, "gets points" for those inputs) must make some other inputs larger ("gets no points" for those inputs).

Cala mite

Due magneti parlano al telefonino. Ovviamente il fatto che sono magneti causa interferenze nella trasmissione. Insomma, si sentono a tratti.

Lottavo nano

Dopo che Biancaneve arriva tra i nani, è lei a cucinare per loro, che non sanno farlo. Ma prima di Biancaneve, chi svolgeva questo compito? Evidentemente un ottavo nano. Non lo vediamo mai, ma è normale, visto che evidentemente sta tutto il tempo chiuso in cucina. E certo non deve essere una bella vita - un'esistenza così non puo' che renderlo triste.

Ed ecco dedotta l'esistenza di Mestolo.