activatedgeek

Fair enough.

I don't quite care about obscure and beautiful theoretical problems. I am looking for new developments that have shown promise to make big practical impact.

For what it's worth, I am no linear algebra definition police so feel free to list what you think is reasonably relevant.

A question for the more informed:

"What are recent developments in linear algebra?"

Perhaps, we can broaden the scope and include numerical/randomized linear algebra.

I come from a machine learning background, and the most recent directions I am aware of are in applications of Krylov subspace methods to very large number of data points and low-precision training (both in the context of Gaussian processes).

"Ageless: The New Science of Getting Older Without Getting Old" by Andrew Steele (2021)

For a very no frills account of the current state of anti-aging research.

I suspect, this may be far too common for anyone who lives a "checkpointing" style of life, i.e. once I achieve this, I will pursue this ambition.

I've realized this sort of checkpoint-based planning is not helpful for the long run because naturally, times change, and with that ambitions change. If you haven't already had a taste of making progress towards the said ambition, and it happens to require a distinct set of skills, it becomes quite a mountain to climb.

> Modern research is incremental.

All research is incremental.

Even though in context it appears to speak to the right idea, I just want to make sure the casual reader doesn't emphasize too much on the "modern" part. This idea of groundbreaking research happening out-of-the-blue is nothing but a pop-sci fantasy.

Every groundbreaking idea, whether it be Newton/Leibniz's calculus, Marx's manifesto, or more recently the deep learning revolution led by Bengio/Hinton/LeCun/Schmidhuber, are only groundbreaking in hindsight.

Ideas take time to come into the shape that appeals to the spirit of the times. Sometimes, they win a lottery (e.g. the Hardware lottery [1]).

[1]: https://hardwarelottery.github.io

This article is odd. I use pre-rendered KaTeX with remark [1], plus their stylesheet on the client side. All of the article seems more specific to MathJax than the title seems to convey.

[1]: https://remark.js.org

Season 2 of the Extremities podcast [1] talks in some detail about this in 6 episodes.

I absolutely love the presentation, and it gets delightfully better with each season. Strong recommend!

[1]: https://anchor.fm/extremities

Perhaps, you may enjoy this collection - "The Overedge Catalog: New Types of Research Organizations" [1].

I think it has quite the variety, and may help you distill a broader sense of where things are headed (and find people to reach out to).

[1]: https://arbesman.net/overedge/

> UN is a joke.

I just finished reading "Washington Bullets" [1]. (Part of) It is a crash course into how and why institutions like UN, IMF etc. are they way they are.

And honestly, I just feel paralyzed after reading the book. These institutions just don't have the clout by design.

[1]: https://www.librarything.com/work/25426761/book/200164537

I don't quite understand the value of using a LaTeX editor in JupyterLab. For most purposes, JupyterLab already supports a subset of LaTeX within the Markdown cells.

Perhaps, the objective is to make JupyterLab the IDE for everything. I just hope JupyterLab doesn't go down the path of becoming everything for everything because honestly, I just don't see why the community's attention should be divided to solve problems that they aren't in the business of in the first place. But as a demonstration that you can run a latex compiler with JupyterLab plugins, great!

I recently started using Texpad [1]. A pretty fair $30/year/seat price point, if you want a local alternative to Overleaf [2]. Their incremental compilation is delightfully real-time; the lag from typing to rendering the PDF is virtually zero.

There are a few unsupported packages which I think are mostly not needed [3]. For instance, the minted package for code highlighting is something that I felt I needed, but then almost never do. The support for microtype package is hopefully coming soon. In any case, the usual fallback is available to MacTex and is supported within Texpad.

[1]: https://www.texpad.com

[2]: https://www.overleaf.com

[3]: https://www.texpad.com/docs/tutorials/tex/typesetters/texpad...

Oh wow! I totally missed all the alchemy and theology studies he was into.

In this particular case, my argument is more along the lines of what this "team of experts" does is a conscious choice throughout.

Black swan events can probably describe a lot of things, but my understanding is that an individual or a team does not really set themselves up for black swan events. It just happens.

I think the key statement to realize here is this,

> Most implausible-sounding ideas are in fact bad and could be safely dismissed. But not when they're proposed by reasonable domain experts.

Off the top of my head, I vaguely remember that the latter half of the careers of many "greats" like Einstein, Fermi, Erdos were riddled with pursuing directions which bore no fruit (to this day).

I don't quite think that crazy bets work on "average". On average they fail. It mostly so happens that crazy bets are on "average" taken by people with a network that can attract able and passionate people, who can see through an idea to its conclusion. This is where experts come in.

Experts often rely on validation by their "community", and by the nature of community dynamics need to upsell. They naturally attract the kind of people needed to execute crazy ideas. But then attracting such talent also has a corrective effect on the originally crazy idea, such that it has a higher chance of succeeding by minimizing the blind spots, the unknown unknowns.

While this article focuses on a niche, tiny wins are a universally applicable concept. I think they are a powerful abstraction.

I recently finished "Tiny Habits: The Small Changes That Change Everything" by BJ Fogg [1], which talks about this in generality. The essence is that behaviors are a product of motivation, ability, and prompts. By keeping things tiny, we avoid the need of very high motivation, the need for very high ability, and devise easy triggers/prompts to get work done. It has been quite fun to identify tiny habits in my daily routine, I've collected quite a few and narrowed down on what works and what doesn't.

This is then applicable to software engineering, product development etc. One of Instagram's co-founder was apparently enrolled in the author's class, and therefore pops up as an example a few times through the book.

I've always ridiculed self-help books. This book does have a self-help flavor, but always concludes with precise actionable advice. By developing a general framework and a mental model around habits (think abstraction in usual software engineering terms), it does feel more controllable. Highly recommended!

[1]: https://www.librarything.com/work/23220746/book/198725553

Vavilov was such an extraordinary individual. While he was imprisoned in 1940 for defying Stalin-era leader Trofim Lysenko for his Mendelian beliefs, his collaborators protected the then largest seed bank in the world during the Siege of Lenningrad [1] by Nazi Germany. Most of those "protectors" died, and Vavilov died of starvation in prison - the same thing he was so obsessed of solving for everyone.

"Where Our Food Comes From: Retracing Nikolay Vavilov's Quest to End" [2] is the canonical read on this topic!

[1]: https://en.wikipedia.org/wiki/Siege_of_Leningrad [2]: https://www.librarything.com/work/7565437

I don't quite agree with this reduction. A good reference textbook is specifically designed to convey a clean linear story of the otherwise ugly conceptual development of research ideas. Notes are personal. Textbooks are a deliberate transform of those notes meant to convey structure in ideas to the average person in the target audience.

I find it funny that someone would go through the pain of undertaking an endeavor as large as writing a textbook, just for themselves. For that, they already have their notes. If you are hinting that writing textbooks (good or bad) has professional consequences, sure. Are they wrong in doing so? I don't see why they shouldn't bear the fruit of good exposition.

Stretching the argument further, you might as well explain almost every action as "people do X for themselves". Kevin Simpler explores this theme in detail [1].

[1]: The Elephant in the Brain: Hidden Motives in Everyday Life (https://www.librarything.com/work/19982533/book/195649617)

I think these notes are great, and Vitaly certainly seems like a great person from Twitter (been following for a while now). I just want to spell out the obvious - the biggest (and probably the only) beneficiary of such structured notes is the note-maker.

The beginners who come in feeling excited that this will be a great learning resource are probably missing the point. Learning happens when you force yourself to create notes by finding structure in the raw text. Notes are extremely personal, and reading someone else's does not have the same emotional connect.

Am I suggesting you stop reading notes made by others? Absolutely not! I am suggesting you rather double down on that, except _always_ make your own notes if the objective is learning. Use the excellent public notes to build your own mental models of what makes for good notes.

Wow! This is the first time I've learned about Adam Curtis, and he doesn't mince any words. This article is a crash course into the trajectory of politics over the last few decades. The interviewer must be commended here too!

As someone who's been fascinated by Richard Dawkins' motivations, it was revealing that as Obama pushed religion outside the White House, the whole Dawkins counter-culture against religion was made irrelevant.

It is, however, interesting that Adam Curtis says that people will find a purpose again with the resurgence of religion (Dawkins has a second chance then? Are we in a loop?). I wonder what HN thinks about that.

And, guess what? Since, the Galerkin approximation requires one to choose a basis that is appropriate to the problem at hand, we now have a deep learning solution too (since neural network learning is essentially equivalent to learning an adaptive basis).

It is called the Deep Galerkin Method [1]. In a nutshell, the method directly minimizes the L2 error over the PDE, boundary conditions and initial conditions. The integral is tricky though, and computed via a Monte Carlo approximation.

[1]: https://arxiv.org/abs/1708.07469

I think I just stopped my train of thought and hit reply.

I meant to point out that Charts.css relies on having <table> elements in the HTML. Vega relies purely on JSON. This may or may not be of concern. For instance, if you care about bots crawling the website for semantic information, then perhaps the Charts.css way of operating on top of <table>s is preferred. A minor point though.

Thanks for this view! Added to the reading list.

Probably as a cautionary note: Reading a decade old book on fast evolving topics may be a bad idea.

Somehow, I don't see the mention of the Vega visualization grammar [1] as a potential charting library. I think it is incredibly well-done, and is generally intuitive in its API. I've mostly used it through the Python bindings called Altair [2]. The good part is that that general grammar carries over to the browser as a full JSON spec that can be used directly in any language that supports JSON and has a Vega binding (which is pretty much all popular languages).

Having said that, I had to unfortunately abandon it because the ad-hoc control with Matplotlib [3] in Python is just infectious. Visual manipulations are far less easy to do in Vega. Being in JSON is also a restrictive though, because it is less interpretable by unstructured bots, where charts.css probably excels by design.

[1]: https://vega.github.io [2]: https://altair-viz.github.io [3]: https://matplotlib.org

On this note, reading Vaclav Smil's "Energy Myths and Realities: Bringing Science to the Energy Policy Debate" [1] has been sobering. I had a similar presumption that this isn't a technological problem but rather of a policy conflicts (and subsequent perverse mechanisms including corruption).

In fact, it is actually a technologically challenging problem. This book is a decade old but I would assume that nothing has drastically changed over the past decade, and the energy demand has probably only gone up. Highly recommended read, if not already!

[1]: https://www.librarything.com/work/10653100

The implosion argument in this article is solely ~based~ triggered by Google's botched handling of AI Ethics researchers. To argue that this will implode the complete AI organization seems like wild stretch. On a normal day, all the arguments padding the article would fall on deaf ears.

So, the answer to the question is no. Instead, AI Ethics research will probably implode. AI research, probably no.

For AI research to implode, major AI research leadership need to be seen leaving. And I am not talking about the corporate leadership but the research thought leadership whose sole purpose is to shield the junior/senior researchers from corporate BS (and who are themselves well-known researchers).

Edit: It is perhaps be more appropriate to have said that the botched AI Ethics researchers handling solely triggered the article.

This article sort of reasserts the same sentiment that the parent comment alluded to. Quoting a couple -

> Some farmers worry whether the reforms might lead to the end of wholesale markets and guaranteed prices.

> Farmers might go from the local monopsonies of the APMCs to the national oligopoly of Amazon-like behemoths.

Pretty cool!

Can someone summarize how this works?

Is this simply a numerical integration of volume and then some assumptions to compute the buoyant force?

For the interested, here is a classic article that goes into more depth: _Practical Markov Chain Monte Carlo_ [1].

[1]: https://projecteuclid.org/euclid.ss/1177011137

Of course. My point was that it appeared like they were actively working on developing new technologies. Usually, a project that is expected to shut down would mostly lose momentum months before. Although, it is possible that this work was done months ago and only came out now due to the slower Nature review cycles.

This is quite a surprise. Only recently, Loon collaborated with seasoned reinforcement learning researchers to build more efficient navigation controllers for the balloons, titled "Autonomous navigation of stratospheric balloons using reinforcement learning" [1]. Although, it would appear that much of the gains come from a simulator that they've been developing for a while.

[1]: https://www.nature.com/articles/s41586-020-2939-8

That's a separate thing. I only meant to point to the heading "As a maximum entropy distribution".