I have decided that weekends is for implementing cryptopals – at least sunday night after a nice bath with epsom salt, candles, incense and cryptopals mood music. I am starting to get faster with rust. My code needs refactoring, optimization, and unit tests, but I really like the rust linter (ale) I am using – thanks folks on the zulip message thread. This site for binary/hex conversion was also helpful in my debugging.
I am extending my RC stay to the full 12 weeks. This gives me some time to r&d my prayer coin once it is released it in the next week or so. It will also give me the chance to bone up on my fundamentals such algorithms and distributed computing, similar to the cryptopals exercises. I can also develop the core concepts of conscious computation.
I finally finished reading James Hillman’s The Dream and the Underworld. Now I finally understand what people mean when they say the ancients had mythology (powerful forces explained as stories of the gods and heros), and the moderns have psychology (powerful forces explained as personal complexes or features of the mind). I also was interested in the idea of a topos of the dream world vs a geography of the dream world and the similar relationship between a depth psychology and a biological study of the mind.
Loosely speaking, in the first instance you treat the mind as a whole with its own logic and landscape, in the other, as a collection of parts to be measured and parsed independently. There is a point that Hillman talks about regarding the inaccessibility of the unconscious. The idea that there area aspects of the mind, and of the dream world in particular, that we cannot access, e.g., that which is hermetically sealed, that which is invisible. We can only access these things obliquely.
I was thinking about this in relation to Machine Learning, since I have been repeatedly binge listening to Machine Learning fast.ai course. What we learn about a model generated by an RNN or CNN (or some other NN), is learned obliquely. How we reason about the speed of a compiler or a computer program is also learned obliquely. These things, machine code and neural networks, are also hermetically sealed. How can we access something that is hermetically sealed? Is it through the aesthetics of the enclosure? The style of the computer language used to generate the machine code, or the data set and training used to generate the NN model?
It reminds me of Deleuze and the fold (and the Baroque). The ultimate hermetically sealed object is Leibniz’s monad (its kinda like an atom with no subatomic particles). How do we communicate with the monad? We have the new concept of the fold! The topology of the interface, the folds, allow stuff to come through because when you look at a curve analytically (with calculus) you will come up with holes! This allows us to obliquely see the light within. Meditate on that – grasshopper.
Some things I learned:
I learned about a new npm package, cross-env, which takes care of setting environment variables across different systems.
I am looking into parsing chess game files to use with my Haskell music program. This will be interesting because I will have to write a Haskell parser, which I feel is a very Haskell thing to do. This is something I have wanted to do for a while, since the 8×8 chess board maps well to the western musical scale.
Michelle wrote a great blog post on how to release an npm package.
Elias posted about cool linux tools.
Someone on zulip posted this course on hacker tools at MIT, which I probably will skim at some point. This is probably the most useful course anyone can take.
RNNs Vs CNN. RNN pays attention to the order of things, such as world always comes after hello. CNN pays attention to the pattern of a thing, such as this is round and this is angular. It is written logic vs pictorial logic. What about the logic of world building or algorithmic logic? What kind of neural net do we use for that. Perhaps that is in the next fast.ai video.