gajjanag

> Those kids of work may even help get you promoted and open up more interesting projects.

Do you have any anecdotes in support of this view? My general understanding is that working on such things (if it isn't part of your explicit job description) is usually met by at best gratitude from other people, but certainly not promotions. And there are definite risks: since it is time spent not on your "actual" job, it can show up in performance cycles especially with a mediocre manager.

Does anyone here have a good source for how fast vector instructions (AVX2 and the like) will compute under Rosetta2 on the Arm-based Macs vs the current x86 Macs?

Very sad to hear. From what I know of Berlekamp's academic work, he always had a good sense of fun in his research. Such researchers are a rarity in my experience, and thus bring a very valuable aspect to the research enterprise.

As an aside, not many in information/coding theory know that he was the advisor of Ken Thompson (of UNIX fame) according to Wikipedia: https://en.wikipedia.org/wiki/Elwyn_Berlekamp

> [...] Many of the most brilliant human beings of that era worked for very modest salaries, and seemed content to do so. I wonder why it is that nowadays the "smartest" people tend to work in tech or finance and mostly pursue wealth. A few possibilities:

This is something in essence that some of us at MIT get told by faculty upon telling them that we want to go into industry. Behind it, though, there is often a fair bit of arrogance: many of them can not truthfully say that they would not do the same if they had to accept an academic position at a "tier 2" place, long postdoc, etc.

That said, the spirit of the point is perfectly fine. My own personal answer is that I have no personal issue with the compensation in academia; the associated "bureaucracy" which is far more severe today (paper/citation counts, and other things of even less significance) turns me away. It remains to be seen for myself how "industrial bureaucracy" compares with it. However, everyone I know has told me that in terms of personal happiness/fulfillment, people who leave are usually much better off.

I mention this as I had a conversation where someone essentially assumed point 1 or 2 rather than the distinct reason above.

Witten's talk at the IAS on this subject may also be of interest: https://www.youtube.com/watch?v=XYugyhoohhY

As others have mentioned Xerox PARC, I also recommend the "PARC counterpart" to this book titled "Dealers of Lightning: Xerox PARC and the Dawn of the Computer Age" by Michael A. Hiltzik. It was a beautiful summer read, full of amusing anecdotes and written in a laid back style while still conveying lots of information about the environment in PARC.

It definitely is not just you. I spent my first two years of PhD wrapping my head around the stats and maths commonly used in ML, and realized that mathematically (as "theoretical" ML is practiced today), most answers are already provided in classical work of statisticians and probabilists. There are many fascinating questions of probability theory and statistics, but most have little to do with AI. In fact, in terms of the biggest empirical success story (deep neural networks), there are essentially no theorems providing a solid conceptual leap of understanding. Mikhail Gromov goes one step further regarding the lack of theory for neural networks (https://www.youtube.com/watch?v=g4Wl3Ggho6k), and provides a fascinating overview of his thoughts in: https://www.ihes.fr/~gromov/category/ergosystems/

I am interested in the points you raise, but also realized that I would not find a good environment for it at MIT in EECS, for reasons that are rather obvious from the article's subtext. As such, the last year or so has been spent in a search for good alternatives in terms of research, and I am slowly finding answers. I am happy to discuss more over email.

Long story short: you are certainly not the only one who thinks that way.

EDIT: added a video link to Mikhail Gromov's actual views for better accuracy.

The funny thing is that MIT already has something called "MIT Institute for Data, Systems, and Society" (IDSS): https://idss.mit.edu/

I wonder how this is going to play with the new MIT Schwarzman College of Computing, and what each entity will choose to focus on.

A funny thing I have noticed about Wikipedia is that although citing Wikipedia itself in an academic paper is still not a good idea generally speaking, the references provided in a Wikipedia article are usually far more relevant than the ones provided in a typical IEEE paper.

It was a shock for me to discover that the "literature review" of an IEEE paper I was reading was essentially useless compared to the one provided by Wikipedia on the same topic. When combined with the fact that the IEEE paper was behind a paywall and the Wikipedia article was accessed within seconds using a search engine, I now know better how to do a literature review...

Why is this the case? Part of what I have noticed about IEEE is that at least 20% of the references provided are simply added at reviewer's insistence to boost their favorite author's citation count (possibly themselves), regardless of what it has to say about the topic at hand. I don't really know how much of this extends to non-IEEE venues. What I can say is that the mathematics papers I have read provide much better references.

So here is a possibly naive question.

From my viewing of Youtube, Google transcodes uploaded videos (typically H.264 stuff) to VP9 only when a certain view threshold is reached, which makes sense from their perspective.

However, I have also noticed that due to the chain of encodes source -> H.264 -> VP9 (latter two available to Google), the VP9 stream is often of noticeably lower quality. Thus, whenever I can, I use an H.264 stream.

This problem will not go away with AV1. In fact, from an archival/local usage standpoint, as others have noted here, AV1 is pretty much impractical due to heavy encoding time increases that will unlikely go away with SIMD as compared to x265 or x264.

As such, from an end user experience point of view, what does AV1 offer that H.265/H.264 do not already?

The Julia programming language uses it.

> Indeed, if the purpose of a PhD is to expand the totality of human knowledge, why was it expected that I would publish 3-4 conference papers to graduate?

This is a question that bothers me quite a bit with respect to EECS (electrical engineering and computer science). In mathematics, one's PhD thesis can often serve as one's first paper (once revised). Funnily enough, the same was true for EECS even at the top tier places in the 1980's - I know of a case at Caltech where the student did not have any publications prior to graduation, and yet went on to become a successful professor.

Is it really true that humanity is producing more knowledge in EECS today per student than it was in the 1980's? I seriously doubt this. In my experience, instead what has happened is that the average "delta" in a conference proceeding has gone down significantly, in spite of the vast increase of words like "novel" and "new" being used, and the ballooning of the average paper length.

In mathematics on the other hand, top quality authors freely admit that a lot of their work can be "implicitly" traced back to the "big names" of the past; indeed this has to be done for conceptual and historical clarity.

I see that you mention Peter Sarnak in your top comment. Peter Sarnak has a a wonderful article at the very end of "The Princeton Companion to Mathematics" on his advice for students. Among other things, he stresses the importance of the "history of ideas", and how it often brings clarity to an entire field.

In a sense, yes. I currently have a Moto G 1st generation that I acquired approximately 4 years back. I intended to keep it for at least a year or 2 more, but through the steady accumulation of such 100 mb losses, I have ~ 500 mb storage left on an 8 gb phone.

Now, with a reasonable OS that should be enough. However, Android insists that it does not have enough storage for app updates, and keeps bugging me with a notification for this. I fail to understand how this is possible given that the app size itself is less than a 100 mb in many cases. Almost every week I am forced to clear the cache in order to perform updates. I believe this has something to do with the swap file - but am not sure.

As a consumer, I feel artificially pushed onto a useless upgrade train. The only reason I am merely annoyed and not infuriated is because I have finally found a legitimate reason (for my needs) to upgrade - a better camera.

I generally agree with this, but here is a counterpoint regarding graduate studies.

From my first hand experience as a graduate student at MIT, a lot of the research produced at MIT generates a disproportionate amount of buzz - essentially identical work (in terms of quality) produced by a student in a "tier 3" school will not generate the same kind of recognition. The reasons for this lie in the network effects; people read and cite people who they can attest to with less effort and scrutiny.

The points regarding the instructors and student body are definitely valid and should not be ignored.

Fiction is in general severely underrated among STEM people today, with many arguing for how it is a waste of time, etc. I was fortunate enough to grow up in an environment where I had good access to it, and a healthy level of support for my interest.

THe thing I find funny about this is that many of the "STEM legends" such as von Neumann, Ulam, Einstein and others had read a lot of classical literature. Einstein himself is quoted saying: "Dostoevsky gives me more than any scientist, more than Gauss." I just make sure I give this quote only when there are no number theorists in the room.

Sure, if one just confines oneself to the "popular" works on a standard YA list, it may be reasonable to conclude that some fiction is not worth reading for a STEM person. However, the classics are classics for a reason after all.

Moreover, it is also not implausible that a discerning eye can find plenty of great worth in modern authors as well, something the parent has done.

You might want to adjust the figure to account for clothing/minimal construction work for shelter - both are definitely fundamental in most parts of the US. One might also wish to add in a minimally functional healthcare system to ensure a reasonably high average life expectancy.

Nevertheless, the point is still extremely valid; I would estimate around 95%.

Update: As in general it is easy to draw overly pessimistic conclusions from this (I have been affected by this myself), it may be of use to point out that positive answers do exist for this issue. One of my personal favorites is the powerful, short book "Man's Search for Meaning" by Viktor Frankl.

Thanks a lot for your work on this issue, and for your references - especially [1].

I had a conversation with a friend a week back where we tried to estimate what is the reduction in economic growth due to the lack of a sane housing policy, and am happy to see that [1] addresses this question.

"Indiscrete Thoughts" is terrific, and there is a lot of gold in its pages.

Of possible interest to those who work in artificial intelligence/machine learning is Chapter IV "The Barrier of Meaning", where Rota records Stanislaw Ulam's thoughts on the matter.

Note also that there is a similar volume entitled "Discrete Thoughts".

For fans of film, essentially all of Agatha Christie's mysteries featuring the Belgian detective character Hercule Poirot have been adapted by ITV into a fantastic TV series: https://en.wikipedia.org/wiki/Agatha_Christie's_Poirot

I find it fascinating that they did this series over a span of over 24 years, with the same lead actor (David Suchet, CBE) as Hercule Poirot throughout!

There is a behind the scenes interview with David Suchet reflecting upon the 24 year experience: https://www.youtube.com/watch?v=FX3ITew9Mpw

"And Then There Were None", which is one of Christie's most famous works not featuring Poirot, has been adapted by the BBC recently: https://en.wikipedia.org/wiki/And_Then_There_Were_None_(mini....

The impact of flying has been very nicely estimated in the freely available book "Sustainable Energy - without the hot air" by the late Prof. David MacKay FRS: http://withouthotair.com/c5/page_35.shtml and http://withouthotair.com/cC/page_269.shtml if you are interested in how he arrives at the estimate.

The book also remarks upon how a (theoretical) revival of ocean liners would actually be worse.

Here is something that I have a need of pretty often, but don't know how to do in things outside email (Whatsapp, Signal, etc): a point by point interleaving of the original message and the reply; for instance in a technical argument.

I could do this in Slack, but unlike email this is not even close to being universally supported.

Disclaimer: I am a graduate student in EECS, and not math. In particular, I am not a professional mathematician.

At some level, none of us is smart enough to understand higher mathematics (in its entirety). Some illustrations (for which references are easy to obtain via a web search):

1. John von Neumann, in his time, said that he understood 28% of mathematics (no idea where he got that specific number from). Given the 20th century explosion of mathematics, largely driven by the professionalization of mathematics and the drive from the nation state, I can't think of anyone today who can truthfully answer > 5%.

2. One of the favorite, semi-secret pastimes of mathematicians is ranking other mathematicians, e.g is so and so a "first rate, second tier mathematician" or a "second rate, first tier mathematician", etc. The funny thing is that this ranking does not end: Fields medallists fall below the "inner circle" of Fields medallists (who e.g won the Abel prize as well), but even they look up to people universally recognized as singular, but who are no longer alive - popular favorites being Euler and Gauss.

A far more modest task is understanding some aspect of mathematics that one finds immensely fascinating. Fascination and enthusiasm are by far the most critical pre-requisites; without them one can't go far. Often what makes a field of mathematics difficult (at least in popular perception), say algebraic geometry for a stereotype, is the lack of familiarity with the language of it. Language takes time to seep into our brains. Some people can do this far faster than others, and hence appear as "geniuses" when they absorb things without much observable effort.

But that does not make it impossible for others; it usually just means more effort over a longer duration.

There is also a pattern to how mathematics operates. Enough exposure to a variety of topics will demonstrate that many famous, deep theorems, are at their core built out of (in retrospect!) simple, natural ideas - ideas so natural that one can come up with everyday analogs of them. In fact, in a highly simplified sense, most proofs at their heart involve what many would call "high school" manipulations (e.g a lot of arguments in real analysis involve what many call a "3 \epsilon" trick). Yes, there are subtleties involved, and a lot of effort sometimes needs to expended to complete arguments, but the same is true of various fields, including programming/software engineering (just look at the evolution of Unix from the original < 10k lines to today's *-nix).

If one wants something concrete that can be appreciated by many on this forum, consider:

1. http://www.math.ucla.edu/~pak/lectures/Math-Videos/comb-vide... - Igor Pak's excellent archive of combinatorics videos; many of which are for a pretty general audience.

2. Federico Ardila's courses, see above for links; also see https://www.quantamagazine.org/mathematician-federico-ardila... If one checks the course websites of his offerings, there are tiny blurbs with some info about the students of his courses. Many come from quite non-traditional backgrounds, and a non-negligible fraction of them end up doing interesting research.

3. "The Princeton Companion to Mathematics" and/or its sister "The Princeton Companion to Applied Mathematics" - these books do a fantastic job of giving a bird's eye view of the essential unity of mathematics, and can be excellent starting points for diving into a topic suited to one's interests.

Lastly, mathematicians themselves do not always understand what they are doing/have done - otherwise generalizations/refinements that take many years to come up with would have happened much faster. This is especially the case when they (in retrospect!) are breaking seriously new ground, and are thus probing far into the dark. Gauss's earliest proof of quadratic reciprocity used induction on the primes; a technique, although useful, is far from most "modern" treatments; and is arguably not the best way to "understand" it. In von Neumann's words: "Young man, in mathematics you don't understand things. You just get used to them."

The vast majority of which don't need, and hence don't effectively utilize a Ph.D.

Most of those jobs can be performed as well, if not better, by people with an undergrad degree, possibly with a master's tacked on + equiv number of years on the job for experience as opposed to the Ph.D.

I went through a similar experience, and spent two years going through the machine learning landscape. Every month or so I would wonder why I still bothered to continue working on it, but still continued as I saw so many smart people around me apparently interested in the topic.

It took a while before I realized that maybe ~10% of the researchers are in it out of genuine interest, and the remainder are simply hoping to ride the wave of success of various kinds (academic, monetary, etc).

And for the 10% who are genuinely interested in the stuff, there is a long history, with many ideas that got explored under various terms as you point out. It is extremely dangerous for a field to lose awareness of the history of ideas affecting it. This is part of what I like about older fields, such as mathematics: researchers at the highest level are very keenly aware of the history of ideas in the discipline, and where modern ``high powered machinery'' can and can't help one on various problems. One can only appreciate the problems and techniques of today's enterprise if one understands what came before.

An analogy I like to think of is that of the problem of alchemy in the middle ages. Turning base substances into gold was a hopeless task back then (and still is right now); yet very intelligent people of the time such as Newton devoted significant amounts of energy into it. From a modern scientific perspective, there is nothing fundamentally impossible about this; stars do it in supernovae. What the modern perspective has given us is the ability to see why the problem is out of reach for now, and why energies are better devoted elsewhere.

My own belief is that 100 years from now, people will be able to articulate similar reasons with respect to the problem of strong AI. A personal favorite in terms of arguments are those of Stanislaw Ulam, as presented in Chapter 4 of "Indiscrete Thoughts" by Gian-Carlo Rota.

In the meantime, there are side benefits to be had from the pursuit of the ``strong AI challenge''; in a similar manner to the great progress in chemistry thanks to the problem of alchemy in the middle ages. Grand challenges serve as a unifying cause, and thus help in achieving remarkable things.

Nice!

I have noticed that there is a lot of high quality mathematical/numerical software coming out of INRIA; and have often wondered of late why we do not have an equivalent in the US. Do you have any thoughts on why this is the case; and what the closest equivalent in the USA may be?

> I guess there is actually an 11th person saying, awesome I can financially support this evil company in the short term and make a quick buck.

Well, unless you belong to the minority of people on this forum with no assets in the stock market whatsoever (either directly or indirectly via a bank account, pension fund, etc), you are part of the system that profits off corporate activity. The only exception is if you traced your holdings in such companies (Wells Fargo is the example given here) and where they are getting used to within reasonable limits (some activity may not be disclosed to you as a member of the public for instance), and divested from all of them. A priori, I would find such a situation highly unlikely.

All that the grandparent commenter did was to allocate/deallocate to get a better return, and is honest about it as you point out. What more can one ask for or expect? Full divestment to within information limits as outlined above?

And as you point out, changing the system is super hard.

> And as disheartening as a comment likes yours is, it's truthful, and I personally find it motivating and inspiring. I guess it's sort of like telling someone something can't be done and that being the catalyst for action. I'm pretty motivated at this time to create an organizational platform for boycotts in a crowdsourcing-esq style without need for funding.

I am very happy to see people with energy for these sorts of things; feel free to email me to discuss more.

> If you don't mind, I guess like a sort of prelaunch feedback, what would it take for someone like you, willing to profit on bad corporate behavior, to instead become a leader for change?

I can give you mine:

1. I need enough savings to live on a "student budget" for the rest of my life in the United States. Until such an amount is reached, compromises will be made, such as what you term "finanically support this evil company in the short term and make a quick buck".

2. I do not wish to do this full time at any stage - there are other things in the world that interest me far more (chiefly scientific questions, usually from mathematics/applied mathematics). For example, if you provided a list of such-and-such companies engaged in such-and-such, and provided a summary of why you consider such a boycott worthy, I would evaluate it. Once upon a time I would have done the research myself; but I no longer have the energy and have a different set of priorities. On these lines, I have found http://www.givewell.org/ and more generally the "effective altruism" movement fantastic.

> With graphics!:

A note: does not work in Firefox as it uses the WebP format. Works in Chrome though.

> A political fix is a necessary prerequisite to the adoption of any technological fix.

Political "fixes" are by nature temporary in the scheme of things. Regimes come and go, governments rise and fall, etc. They can possibly serve as useful incentives in the short term to drive certain changes, assuming there is sufficient faith in the longer term stability of the political "fix". Such faith has been shaken due to various reasons.

Technology, on the other hand, does not regress so easily IMHO. A technological advance that is sufficiently compelling - say fusion power for the sake of a dramatic illustration, can make previous "dirtier" technologies obsolete, and price them out even in the presence of improperly calculated externalities.

> that I don't see a future where they are unseated.

The way I see it, consumers (at least right now) are free not to buy into the hype and consume whatever it is that these companies produce. So if prices are too high or quality too low, a new company can enter in.

The real danger I see is the possibility that essentially all consumers will be forced to pay for this stuff through some regulatory capture, in a similar fashion to the state of the healthcare industry in the US.

This is an astute observation.

The world leaders during WW2 recognized this - basically, when critical work needed to be done, they knew whom they needed and roped in a large number of the leading "dreamers/thinkers" of the time to work on the Manhattan Project.

The thing that saddens me is that it is often conflict and animosity that awaken this recognition - other examples include the Cold War, artillery calculations in WW1, etc.