wolfgangK

About words embeddings, the №1 example is the famous King - Man + Women = Queen This works nicely in the vector space but fails to make a visual impression when projected on 2 dimensions. Neither with ACP, nor MDS ot t-SNE in my experience : https://bhugueney.gitlab.io/test-notebooks-org-publish/jupyt...

(← JupyterLite Notebook doing words embedding in the browser : don't try to run this on a smartphone !)

Does anyone know how to nicely visualize the poster child of words embeddings ?

I really wish there was a name (that I knew of) for this logical fallacy of using precise quantities of meaninglessly confused categories. What do you think "farmland" is ? Our ancestors who knew how to raise cattle and farm cereals for millennia were surviving on the verge of starvation : why do you think that was ? Were they so dumb they'd rather die of hunger than convert some of their grazing pastures to cereal agriculture ? Or some "farmland" can be used for grazing but not for agriculture, maybe ? Same for proteins : all amino acids are not equal, a a pound of soy protein won't have the same effect on your body as a pound of meat proteins. And don't get me started on nutrients and their bio availability (e.g. iron and B12). As far as CO₂ emissions from cattle, where do your think the C comes from ? The grass/food that cattle eat, what would happen to it if let to decompose ?

The dirty secret of education is that no school outperform nor underperform much other schools when adjusting for differences in student populations :

https://freddiedeboer.substack.com/p/why-selection-bias-is-t...

If I may hijack a bit this thread, I am currently looking for an optimization solver for a specific goal : I wish to constraint the values of variables so that tuples of variables are unique within sets of tuples.

Cf. https://stackoverflow.com/questions/71878354/constraint-prog...

For now, I could not find anything satisfactory, so any recommendation would be most welcome. Otherwise, I guess I'll have to generate a variable for each tuple and use and all_different (e.g. https://cpmpy.readthedocs.io/en/latest/api/expressions/globa... ) on those.

I'm sure most people would want that, but I don't think that we know how to "make people" change. Much in the same way that we don't know how to have school "make children" more academic, we don't know how to make inmates less violent / impulsive / thrill-seeking (except by the inexorable passing of time). I'd be very happy to be corrected, tho.

OTOH, giving proper (i.e. often involuntary) treatment for violent mental illnesses would for sure improve the situation.

I would have expected the Clojure versions of types to inherit from the Java versions. Any reason not to do that ?

Do you actually thing that the anti-IQ crowd is mad that the nazis (notoriously anti-IQ) lost WWII and that is the reason they are so touchy about this ?

It makes no sense to posit that IQ differences between population have a different cause than IQ differences between individuals. Adoption raises the IQ by at most 5 points ! What environmental difference could have a much more dramatic effect than that (between population IQ difference are up to 15 pts) ?

Most interesting ! I love the use of Pyodide. It reminds my of https://basthon.fr/about.html . What is the libraries situation ? I'd be most interested in pandas, seaborn and SciKit-Learn to contribute teaching materials. Also, I have had the time to look into the architecture, but the best use of the Open Source nature of such project would be, imho, to have both a repository of teaching materials and one or more courses assembling a selection from the teaching materials repository. So that people could easily fork the repository to reuse some of the teaching materials and create new teaching materials for their own classes and contribute their new teaching materials upstream to you. Also, I haven't checked into the i18n / l10n situation of your project, but it would be nice to have the infrastructure to easily provide translations (I would be happy to contribute in French). Kuddos and best regards !

Has education ever been ? Can it ever be ? I'm not sure it makes sense to hope that anything could prevent offspring to be more similar (and have more similar lives) to their parents that to random member of their species. Gregory Clark is very sobering in this regard :

http://eprints.lse.ac.uk/110873/1/Working_Paper_67.pdf

EDIT: See also https://edi.opml.co.uk/wpcms/wp-content/uploads/2021/02/Kyrg...

Nice ! Now I just need to be able to have org mode source blocks for that .

Being a teacher myself, I understand this policy. Too bad, though, I'd have dearly loved to read the python implementation to rewrite it in Clojure. I think the language could allow for something at least as concise be maybe cleaner.

Or maybe there might be an other confounder than skin color ? Maybe SES status of drug addict influences the way they finance their habit You might want to compare the impact of crack in black neighborhoods on violent crime and the impact of opioids in white neighborhoods on violent crime. Assuming the justice/police system is actually concerned about violent crimes, Occam's razor shaves off the racist assumptions pretty nicely. Also cf. Locking up our own https://www.nytimes.com/2017/04/11/books/review-locking-up-o...

> Thompson sampling itself is not difficult (specifically, 6 lines, as I'm looking at my Clojure code),

Going through "A tutorial on Thompson Sampling" and being eager to hone my skills in Clojure, I would be very interested in reading this code if there would be a repository online or a gist you could share with us.

Best Regards.

> IQ is NOT a measure of a thing that exists.

I don't understand your concept of something "existing" or not. Do you believe that a "speed" exist ? That a "center of gravity" exists ? That a "temperature" exist ? People who say "IQ is what IQ tests measure" sound to me like they could say "temperature is what thermometers measure". temperature does not "exist" it is molecules movements all the way down, yet it is a meaningful/useful concept. Same goes for IQ or "g factor".

I consider myself as leftist as possible with a functioning and informed brain. Yet I know that "left leaning" and "right leaning" are mostly biases that we use to orient ourselves in a world too complex to be fully understood. EDIT : I think that everybody should read http://righteousmind.com/ from J. Haidt.

Wrt to whether differences are biologically based or not, I'm convinced that one day, we'll look back at the Nature vs Nuture debate like we do with the Particule vs Wave debate for light. The sooner the better.

As to which part was so offensive, it's akin to wonder which spells got the witch burning at the stake. Witnessing so many smart and compassionate people join the angry mob really scared me : I have no reason to believe that I would not do the same if the right buttons were pushed.

I hope that he also sues media outlets for slander. To sum things up :

  - fairness & lack of prejudice wrt minorities/women : goal
  - minorities/women % in the workforce : metric
  - diversity hiring : https://en.wikipedia.org/wiki/Campbell%27s_law because https://en.wikipedia.org/wiki/Berkson%27s_paradox .
  - without diversity hiring, differences in women vs men applicants pools, which can be explained by differences in *distributions* of interests, have no impact on distribution of skills in the workforce
  - women are talented enough so that they just need to be *attracted* (e.g. better work/life balance) rather than pity-hired

This is my takeaway from the "screed". How could it be framed as «anti-diversity», much less «alt-right» ? How could saying that «women are, on average more people-oriented and men are more thing-oriented» be framed as «women are inferior, biologically incapable of coding» ? I hope that people get punished for such a dishonest character assassination !

The problem is that reverse discrimination does justify the discrimination they [cl]aim to fight. If HR considers the results of my technical interviews to be biased for the benefit of people who look like me and compensate accordingly, I then have to discriminate as I'm supposed to in order to achieve overall fairness of the recruitment process. How is that supposed to reduce biases ?

Reverse discrimination is the easy way, but it is actually counterproductive in term of reducing biases. As case of https://en.wikipedia.org/wiki/Campbell%27s_law imo, of optimizing the easy metric of enrollment ratio instead of the hard goal of fairness.

> say they are inferior in some respect

I do not believe that the author of the memo ever wrote anything like that. I'm not sure that you should equate all research scientists who published on gender/sex differences with nazis.

Exactly !

But 'intertwined' means correlation, not causation. Being a woman is correlated with not doing software dev. in a tech company. It does not mean that tech discriminate against women, because correlation ≠ causation.

> Sure, if you define diversity as having nothing to do with gender,

I would have sworn that this would be "Not being sexist 101", but apparently words do not mean what I think they should…

Re: Ukip and racism, there is a lot of political acts to provide context that can invalidate (or not [0]) their statements. As for this guy, his memo should be considered on it's own (in good faith) for lack of outside context about the author.

I do not understand how what he actually advocates could be considered discriminatory. It seems to me that it would indeed "disproportionately affect particular demographics", but in a good way by making sure that no https://en.wikipedia.org/wiki/Berkson%27s_paradox could induce any skill/gender bias in the workforce.

[0] I have no particular knowledge about ukip

> By stating that he's OK with discriminating by gender,

Actually, he wrote exactly the opposite.

There is no way around https://en.wikipedia.org/wiki/Berkson%27s_paradox

Fond memories of programming my Amiga to scrap pages from the Minitel to fill up up database instead of doing data entry by hand, as ha been expected of me during my first intership…

Point-free programming would be extremely useful if the composition primitive allowed introspection. You would then compose functions that would have run-time homoiconicity as you would be able to edit the "ast" embedded in the point-free function.

I use this to modify the parameters of the composite geometric transformations of fractals. This allows me to generate animations for any kind of fractals without having to reimplement the animation code for each fractal : https://scientific-coder.github.io/Playground/2017-03-20-fra... (cf stepify multi-method). This would be useful for web servers in Clojure as we could have our cake and eat it too : while currently one has to pick either easy with ring handlers that are simple function compositions and vectors of interceptors for pedestal that can be modified at runtime.

Always nice to see some clojure code, even more so when to code is interactive. However, I would not have coded it that way. Call it premature abstraction if you will, but I think, especially in a language guiding you toward purity and immutability like Clojure, that one should avoid "doing" things until the last moment. So instead of drawing triangle, I'd compute them, and only draw the computed triangles as as sequence of polylines. The added benefit it that you can them compose any other transformation on the resulting triangles.

I have not found the time yet to actually write the blog post, but my take would be more like :

https://scientific-coder.github.io/Playground/2017-03-20-fra...

Your post is very condescending but also very light on facts. To someone successfully using C++ metaprogramming you just sound like you failed to find how and when to use it. (I do agree that it can get ugly wrt error messages and syntax as neither the compilers nor the language were designed with metaprogramming in mind.)

I would have sworn that Fortran was compiled.

I believe it to be a metaphor on that corporate culture clash between "do no evil" and "cthulhu was a hippie".