n-Category Café

There aren't big population-level differences, aside from basic cosmetic differences due to genetic drift [0]. Humans act as one single species, with one single pedigree [2], lacking any selfish genes which speciate us further.

Genes within a single species' gene pool follow a Poisson distribution because they are randomly permuted [1][3][4], assuming the pool is large enough. This means that often, some genes will be popular for no good reason; that is, these popular genes are cosmetic. When, for example, most people in the world follow one of several popular skin-coloration patterns, that does not therefore imply any sort of fitness function is selecting for skin colors, because the null hypothesis that they are randomly permuted has so little problem explaining how they could arise without any effect on fitness.

These questions aren't dangerous because they somehow subvert existing institutions, but because they're a direct introduction to ethnofascism via eugenics. We should roundly not just reject them based on empirical observations and data, but also based on theory and modelling; this is because we can then use contraposition to flip around various deductions. For example, people often ask whether "race" affects "intelligence" and point to survey data which shows correlations; using contraposition, and understanding that "race" is a social construction, we can refute the idea that "intelligence" is innate and suggest instead that it is tied to socioeconomic status…

The progress in n-category theory is largely about that. Instead of just studying objects, operations and equations, like "a group is a set g, together with multiplication + certain axioms" one studies objects, relations between objects, relations between relations etc. To study this up to level two, you visualize objects as regions in 2d space, relations between objects as lines between them and relations between relations as points where lines meet. Equations are then equalities between such pictures. If you look at those equalities as a dynamic process, which starts at time t=0 with one picture and time t=1 with the other, you get a picture in three dimensional space, which you can visualize as a movie (to be mathematically precise after the choice of a morse function you get a movie). One of the easiest examples would be the zig-zag identity. See for example
https://golem.ph.utexas.edu/category/2006/10/classical_vs_quantum_computati_4.html
and scroll down.

One person who has completely gone overboard with that is Jacob Lurie, he has practically single handedly written down a foundational work of higher algebra and algebraic topology, which essentially precisely corresponds to what you are asking. How does mathematics look like if we weren't constrained to 2d paper to write down equations. Of course to actually write down those things you have to somehow project the equations to several essentially 1d sections or invoke very abstract methods to reason about them. Of…

I have already explained in the post above why optimization is very a much a part of machine learning. Now, you say that that classification and optimization are two different things and that is true. But really, it can be more fruitful to look at supervised learning as a special case of unsupervised learning [1]. It is best to seek the most general framework from which to understand things as it leads to a deeper understanding, broader applicability of concepts and easier cross fertilization across fields.

For example, understanding the spectral theorem makes SVD (hence PCA) and the DFT class of algorithms much clearer. Understand the notion of Lp-Norms, convexity, adjoints, loss functions and regularization and a whole bunch of seemingly different algorithms collapse into facets of the same thing. Hook it up to automatic differentiation then some optimization algorithms and you can write anything from Neural networks, SVMs, regularized logistic regression to Non negative tensor factorization in a few lines. You stop making arbitrary divisions between classification or optimization. Much the same kind of collapse can be done for the dual [2] notion of probabilistic algorithms by thinking in terms of graphs, simplices, parametrizations, families and conjugacy.

The best thing from all this is you stop thinking of which algorithm should I use and start thinking of what do I want to do? What is the best mathematical model for this? What would really be great would be a…

Huh? Nature Publishing Group, where Rowan Manning works, makes its profit margins from extorting scientific libraries rather than superior code.

http://golem.ph.utexas.edu/category/2010/06/boycotting_the_nature_publishi.html

Boycotting the Nature Publishing Group

The Nature Publishing Group publishes not only the
prestigious journal Nature, but also many others. When this
company bought Scientific American, it raised the
institutional subscription price seven-fold. Now they are
insisting on quadrupling the fees for 67 journals to which
the University of California subscribes.

Right now, we pay them an average of $4,465 per year for
each journal we subscribe to. After the increase, this
would soar to $17,479 per year. In response, the University
of California is considering a system-wide boycott of the
Nature Publishing Group — for example, cancelling
subscriptions to all their journals.

Rowan Manning works for a gang of monopolistic rent-seekers who currently have the scientific community over a barrel. His salary comes from a closed-access model which is soon to be obsolete. Thus "cutting-edge" technology is actually highlighted as a fail!

https://github.com/rowanmanning/joblint/blob/master/lib/rule/expectations.js#L6

Article was from 2008, Wired did a follow up story a year later...

http://www.wired.com/dangerroom/2009/10/impossible-drive-designers-dream-flying-cars-stealth-missiles/

----------------
Last year, professor Yang Juan of the College of Astronautics at Northwestern Polytechnical University (NPU) in Xi’an was happy to confirm that they were building an Emdrive which would be tested by the end of the year. But following the publication of this news in Danger Room, the situation changed. I was informed that the publicity was very unwelcome, especially any suggestion that there might be a military application. (Yang had previous published a study on the use of plasma as a weapon against low-orbiting satellites. [.pdf]) No further information has been forthcoming, and no Chinese papers have been published on the Emdrive, though Yang has recently published work on (unrelated) microwave plasma thrusters (.pdf).
----------------

Nextbigfuture did an article stating the tech was sold...
http://nextbigfuture.com/2011/03/controversial-emdrive-claims-to-have.html

A few responses from physicists...

http://scienceblogs.com/sunclipse/2008/09/the_emdrive_story_or_how_to_pr.php

http://golem.ph.utexas.edu/category/2006/10/new_scientist_reacts.html

So, here's the thing I'd like somebody to explain to me. Telegram encryption spec is published, and the client is open source. This means you can verify that the server is following the spec by creating a clean room client implementation. You could even create a green field server implementation if you wanted.

Telegram regularly has contests to break its encryption with a reward of 300K USD https://telegram.org/blog/cryptocontest

If it's not secure, then surely people would be cashing in on that sweet money. So, why is it that we constantly see articles talking about how insecure Telegram encryption is, but nobody is showing a proof of concept attack or collecting the prize?

Unless somebody puts their money where their mouth is and shows an actual exploit with code, it seems like pure FUD to me. On top of that, it appears that attacks on Telegram often come from people associated with Signal in some way. Signal is endorsed by NSA who have a history of promoting weak encryption that they have found backdoors into. I hope everybody still remembers this debacle https://golem.ph.utexas.edu/category/2014/10/new_evidence_of_nsa_weakening.html