covidacct

Xi Jingping consolidated power by rallying people around the flag of anti-corruption. Same with Putin. Obviously, neither of those people is anywhere close to non-corrupt.

Don't rally around anti-corruption. Rally around accountability, oversight, openness, and balances of power.

Doing things right is harder than calling out others for doing things wrong, and plenty of horrible corrupt people are all too happy to call out wrong-doers. Support people who do things right, not people who complain about others doing things wrong.

OT, but does anyone know where the "learnings" phrase in corporatese come from? I noticed it five or six years ago and now it's everywhere. The first time I heard it was from a German VP and I assumed it was some sort of odd transliteration or maybe a Borat reference, but that hypothesis is obviously invalidated.

Seems tautological, especially for a platform biz. Companies that don't "aggregate demand" fail because "aggregating demand" is just a fancy way of saying "have a critical mass of customers". It's tautological because companies that "succeed" by aggregating supply eventually also aggregate demand because... well, because companies without customers don't "succeed". So, if you "succeed", you've aggregated demand, regardless of whether you were able to aggregate demand by aggregating supply. Teasing out the causality is hard, but it'll always be possible to post-facto confirm the hypothesis that successful (platform) companies aggregate demand.

The substantive conjecture behind his tautology is about the economics and business model of SDC-as-ridesharing. Seems like a moot point to me. A working SDC will be enormously profitable regardless of whether the company that builds it gets to own the platform. Plus, the big auto cos might be able to afford to not win the SDC race, but they certainly can't afford to not even play the game. A lot of the money spent on SDCs is seen as hedging, not as serious plays.

Is there some realistic path for me to escalate?

Is the instructor an ad junct? Those are cheap and replaceable. They make maybe ~1/10th of what your average FAANG engineer makes because the labor supply is extraordinarily over-saturated. Enough complaints will probably get them fired, if that's what you're looking for, but realize you're probably putting someone with no prospects and no savings out on the street and they will almost certainly be replaced with someone who gives about as many shits as you can expect from someone with a PhD who makes 30k a year. That tuition you're paying isn't flowing to the people teaching the courses.

If the instructor is an actual professor, then probably there's nothing you can do.

BTW, let your kid at least try their hand at navigating this on their own, even if you provide oversight. Barely competent people doing a poor job is not unique to pandemic times and is definitely not unique to higher ed. Knowing how to use a bureaucracy to route around an incompetent person not doing their job well is a valuable life skill. Plus, everyone involved is going to be more sympathetic to an articulate student than to a "helicopter parent".

I'm mostly on board with the critique of university rankings, but I really don't buy this specific critique. It's a strawman.

Here's what they did: they took the top research universities and the top liberal arts colleges. They then observed that the rankings can't distinguish between the top institutions within each class, and that you can also flip which of those two classes is preferred.

But, in what sense is it even meaningful to compare Amherst College and Harvard [1]? Those are just enormously different types of institutions. It'd be like creating a ranking of "best cars" that includes the top 5 sedans and the top 5 trucks, and then observing that you can jiggle the rankings to get anyone on top. Is a Toyota Corolla better than an F-150? IDK. Stupid question. US News and World Report, for all its problems, does at least get this much right! they break down institutions by "type" and then rank within type.

Additionally, just because rankings are noisy and easy to game locally doesn't mean they are inaccurate or easy to game globally. Two institutions within 10-20 slots of one another are probably pretty similar and rankings aren't particularly helpful / are easy to game. But the #100 liberal arts college is probably not as good as Amherst, and the #78 National University is probably not as good as Harvard, and no amount of gaming is going to change that.

Rankings are indeed noisy and inaccurate and easy to game. But this particular article is not a compelling demonstration of that fact.

[1] For non-US readers: Amherst college (not to be confused with UMass Amherst) is in a class of peculiar institutions that are fairly unique to the USA as described here: https://en.wikipedia.org/wiki/Liberal_arts_college TL;DR: they're basically the diametric opposite of super-charged research universities like Harvard.

They accept the risk and willing to trade the cost with the benefit.

This is only true at the very beginning of a smoker's habit.

I know a lot of smokers. Exactly zero of them are "willing" to trade the cost for the benefit, for any meaningful definition of "willing".

Addiction is hard. Many addicts rationalize, but few truly "choose".

Counter-point: The "everyone needs ridiculous amount of driveway space" rule existed in a suburb I lived in. We had exceptionally wide streets, but parking any car in the street for any amount of time was strictly forbidden. All of my neighbors complained about how "ghetto" our street looked when my parents visited and parked in the street. People really actually hated the idea of anyone parking in the street, even for a weekend, and so we needed driveways that were large enough to handle the 3 days of the year when we had a couple extra cars around.

That rule was dumb and unnecessary, but was also not a "pointless regulation", it was not "regulatory capture" by the Big Driveway Cartel, and it definitely wasn't a response to a legitimate need for more parking (again, super wide streets that sat empty). It was just a reflection of the deeply held convictions of an extraordinarily parochial population.

Roads bridges

See https://en.wikipedia.org/wiki/Public%E2%80%93private_partner...

healthcare

The large healthcare providers are already run like private businesses, because they are private businesses.

education

Please no.

This is a favorite pet issue of the rich and powerful. Their ranks are littered with failed school reform activists. Sometimes those folks even have the self-awareness to realize they failed, e.g., the Gates Foundation. More often, they keep on proposing the same old ideas that are continuously tried and fail, while sending their own kinds to expensive private schools or public schools in the priciest zip codes in the country. Schools that are, aside from their wealth, operated in a quite traditional fashion.

If you look at education outcomes by country, there's really only two things that stand out as invariants that separate education systems that work from those that don't:

1. degree of heterogeneity in schooling (school choice without some strong equalizing effect on quality just rearranges the deck chairs on the titanic), and

2. level of respect for educators (pay != respect).

The "rich people can fix this" attitude is truly both a symptom and a cause of the biggest issue with America's education system: teaching is a low-status job (even places where it's well-paid).

When it comes to education, Americans respect literally anyone's opinion more than the opinion of educational professionals. Hell, even random people who were literally born into wealth get more voice than the professionals who spend their lives in the classroom. If that doesn't change, nothing else will.

But you can't just move your entire country onto solar power

If oil consumption enters extreme secular decline and SA doesn't transition to solar, then SA will be a horrendously poor country powered by oil.

If oil consumption enters extreme secular decline and SA does transition to solar, then SA will be a horrendously poor country powered by solar.

Private capital, during the same time frame, gave us google search and maps, smartphones, movies on demand and god knows how much other stuff

I was perhaps too coy.

https://www.nsf.gov/discoveries/disc_summ.jsp?cntn_id=100660...

I'm not even sure what to say about maps.

Here's one teeny tiny example. They made a smart investment, to the tune of billions of dollars, in computer science. And not just for the past 30 years, but continuously for the past 100 years.

That investment included a modest (6-7 figure, depending on how your count) investment in a project on digital libraries years before private capital caught on. The private investors were rewarded handsomely for their investment in that digital library, but we, the taxpayers, were not.

Federal and state governments are different.

From my original post:

Governments DID see this coming, both federal and state. California, at one time, was prepared for exactly this crisis. The 2008 financial crisis wiped out state budgets

I expect this essay to be the target of criticism. Here’s a modest proposal to my critics. Instead of attacking my ideas of what to build, conceive your own! What do you think we should build? There’s an excellent chance I’ll agree with you.

What we need to build is the capacity to make investments in building backstops against low-prob, high-risk, collective risks.

For almost all of those risks, the problem isn't "what to build", it's that we know what to build and keep failing to build it (or we build it just to burn it down again whenever the next crisis hits). The problem is building sustainable capacity.

Here's my idea. Tax the living hell out of capital. Especially venture capital leveraging decades of gov't investment. Earmark the money for disaster prep.

One possible implementation: gov't gets 50% stake on every patent traceable to any federal grant and every company whose founders were funded through federal grants. I don't think that's insane. Y Combinator takes a 7% cut for 150K. The median NSF grant is substantially larger, and NSF grants are "tiny" compared to other grants, and most of that work happens in places where the funded employees can live comfortably on 25K/year. We would expect an even larger cut than 7% for substantially more investment (often millions) and all of that in low CoL areas. But Gov't gets 0%.

Governments DID see this coming, both federal and state. Governments DID build what was need to respond. California, at one time, was prepared for exactly this crisis. Same as France. The 2008 financial crisis wiped out state budgets, and then the fed govt suffered 6 years of gridlock and austerity driven by owners of capital. Government preparations for this substantial and real risk DID exist, but were liquidated so that the rest could fit in a bathtub to be drowned.

None of that would have had to happen if federal and state governments had a non-trivial stake in FAANG, or even just G.

This talk is an amusing aside. I think the other posters are bit harsh for a talk that's clearly meant to be taken in that way.

If you're interested in actual psychology of programming languages, there is a small but growing research community at the intersection of programming languages and software engineering. The focus is mostly on usability research a la human-computer interaction, but there's also work on community psychology, etc. applies to programming languages and software engineering. See, for example, https://ieeexplore.ieee.org/document/7503516

Corporate debt

I implemented neural networks before the advent of the good python frameworks. It sucked. And CNNs existed for decades before AlexNet. Honestly, the software and hardware engineers are the real heroes of the deep learning revolution.

By way of analogy, David Heinemeier Hansson didn't invent the webapp or even the MVC design pattern. But Ruby on Rails did change the way webapps were built, and enabled a bunch of stuff that wouldn't have been possible otherwise (or, at least, would've taken longer and been more expensive). Lots of websites were built because of Ruby on Rails that probably wouldn't have been built otherwise, even if people would've kept on doing the web thing regardless. We can say the same thing about lisps, about linux, and about a lot of other software infrastructure.

Any high schooler who's capable of learning python and can afford a gaming desktop can build and train a neural network. That's pretty amazing, and definitely wouldn't be the case without computation graph frameworks.

I've never worked for Google or with Jeff, and I'm not a huge fan of the ad tech industry, although I don't understand why either of those things should matter.

Yup, I absolutely agree. Almost all big leaps in software engineering and applied computer science come from building a powerful and simple abstraction. Powerful and simple abstractions are surprisingly difficult to get right.

It's a huge amount of code, hidden behind the tensorflow import statements. It's common to credit GPUs for the rapid spread of deep learning, but good GPUs were available for quite a few years before deep learning really took off. As someone who wrote * a lot* of OpenCL code, including my own python wrappers, I'm fairly certain this code would be thousands of lines without a computation graph framework library. These frameworks are really amazing pieces of software engineering and deserve some non-trivial fraction of the credit for the rise of deep learning.

If you want to know what the next hot thing in software engineering will be, just pay attention to whatever Jeff Dean is doing.

The ability of workers to capture their own productivity gains is a lower than it's been in a century. This crisis will push the second derivative lower.

I've never understood why high-quality chairs were ever symbols of excess. Ask anyone with bad sciatica how much they spend on PT a year. It's probably several multiples of the cost of a good chair. Then ask how much more they would spend to make the pain go away. It'll probably be some nontrivial fraction of their net worth. Bad back pain really takes the joy out of life.

Office people who make fun of good chairs are the whitecollar equivalent of shop/factory people who make fun of hearing protection.

The idea of defining a set of transition rules and then analyzing the properties of some closure (e.g., think reflexive and transitive) of those rules. For transition rules of various orders, expressiveness, etc. And then stepping back and realizing that's what you're studying and generalizing it by thinking rigorously about the relationships between those systems and so on. That's really pretty much then entire modus operandi of a huge chunk of PL research, and there's a ton of mathematical and actual technology built up for doing so. The algorithms that are core components of Stephen's "standard library" for this project scratch the surface.

The question you're asking requires non-trivial effort to answer precisely because "emulation cone" and "rulial space" are never quite all the way defined, and the question being asked in terms of these definitions is also left a big vague.

Emulation cones go by various names, but perhaps the most common is the (bounded) reflexive and transitive closure of a reduction rules of a system. Another common name is the (bounded) reachable set.

Rulial spaces, by which I mean the particular ones Stephen seems interested in toward the end, are higher order term rewriting systems or higher order syntax. But actually, rulial space is used throughout the text in a much more general sense. I'd consider even very canonical results from PL theory, e.g. confluence of rewriting, to be non-trivial observations about a particular rulial space.

The reason for giving (or at least very vaguely hinting at) a definition for rulial spaces and emulation cones is to talk about foliations and then expressiveness. There's some connection between foliation and bisimulation that's difficult to exactly nail down, because nailing it down requires a lot more precision about the exact sort of (emulation cones we are interested in and for which) spaces. The connection between expressiveness and complexity hierarchies is immediately obvious, I think, right?

One frustrating thing that I often find is that much of Wolfram criticism is non-specific and as it's impossible for me to bucket Wolfram I can't bucket his critics either because they tend not to provide enough detail or clarity.

Oy, no good deed goes unpunished :)

Look, I get why it's frustrating.

But, really, there's a reason that rule #0 of technical writing is to define terms before using them. The reader can only do so much.

To avoid redundant threads: https://news.ycombinator.com/item?id=22871806

Oy, I've been dreading having to answer this question since I pressed "post" :)

I've decided that I do not have the time or interest in writing the Related Work section for a paper-length blog post touching on an enormous number of fields, some of which I know well and some of which I haven't thought about in a decade. (As an aside, one real and substantive problem with trying to build a research program without taking the time to share a survey and comparison with related work is that you'll have difficulty communicating with others. It will then take extra effort on others' part to build up a knowledge base. Surveying and comparing to related work is hard and thankless but important work. It's not about credit, it's about building up a shared knowledge base.)

However, I can spare an hour or two and take the time to flesh out one or two in order to demonstrate what I mean.

So, I'll make the following offer: is there any particular excerpt from Stephen's blog post to which you would like the CS/PL/info theory analogue? Or, would you prefer me to pick a particular sentence and identify the body of work that explores that question and the major results of that body of work? (I will take this opportunity to emphasize the "about the non-physics side of things" portion of my original post.)

I'm going to link to this thread in other places where people have asked this question instead of monitoring 3 or 4 different threads going forward. I'll do my best to occasionally monitor this thread for requests and do my best to reply. FYI, I probably won't get around to answering more than one reply until the weekend.

The substance a good dsl for implementing rewrite/string substitution systems, as well as (likely high-quality) implementations of algorithms causal invariants, knuth-bendix completion, graph isomorphism, etc. [1]. A lot of that pre-existed the Wolfram Physics Project. The new thing appears to be:

1. a rebranding with improved documentation / learning resources for using those features (hey, documentation is hard! discoverability matters!), and

2. a set of really impressive rendering algorithms for visualizing these sorts of systems. (hey, visualizing mathematical objects is quite substantive! I'll happily pay good money for "low-effort pretty charts"!)

I addressed in the companion thread why I don't think this is particularly likely to result in "a new type of physics": https://news.ycombinator.com/item?id=22866284

Mathematica is good software. Pre-wolframengine I did shell out the $2K for Mathematica because it contains best-in-class implementations of some important algorithms. A high-quality reimplementation that matches Mathematica's performance would take me several years (in an area where I have a phd). Wolfram Research employs smart folks and lets them spend their time writing really high quality code, and the only way to access that code is by calling Mathematica's built-in functions.

But just because the goal is perhaps a bit grandois and the presentation definitely breathless doesn't mean there's "nothing of substance".

[1] https://www.wolframcloud.com/obj/wolframphysics/Tools/guide-...

Sure. The problem here is that exactly the ideas he's proposing to explore have already been explored. I've slightly edited my previous comment to point this out.

The problem, in the very particular case of this blog post, is that the cost for lacking intellectual humility is spending time reinventing other people's wheels. And those wheels won't get him as far as he thinks they will. We know because they've already been built by others.

Maybe. I doubt it, though. There has always been substantial cross-talk between CS/information theory and Physics. Even through the 1990s it was difficult to be a computer scientist without eventually coming into contact with a non-trivial number of physicists. Especially in industrial research labs. Bell Labs, PARC, and IBM Research were full of physicists. Bell Labs and PARC are dead, but AFAIK IBM Research still has a bunch of physicists and the newer kids on the block (Google Research, FAIR, Deepmind, Microsoft Research, Intel, AMD) also have a share of physicists.

Besides, Stephen's approach here is to ignore 15-20 years of research from various CS sub-communities; his best case scenario is spending a decade reinventing that wheel. The problem with cross-talk that isn't "humble on both sides" is that it's either a) a waste of time because one side's ideas aren't that important, or else b) a waste of time because one side has to reinvent the other wise.

After a few iterations, I think I understand the point of this piece. It was a bit difficult to hone in on, though.

The article suggests that working out the theory of something like rule production systems, and then figuring out how that theory relates to existing insights from physics, is the best path toward a Fundamental Theory of Physics.

My primary source of skepticism stems from the fact that the theory of rule production systems is not exactly a new area of study. It's been well-developed at various points in time and from various perspectives by the theoretical CS, programming language theory, automated theorem proving, and mathematical logic communities. That theory addresses most of Stephen's "big difficult questions" about the non-physics side of things. For example, his "emulation cones" are a new name for a very old and extremely well-studied idea. The term "rulial space", similarly, is a new name for an idea that's well-developed in programming language theory.

I sympathize with Stephen. In fact, he sounds a bit like I did early in my scientific career. Unfortunately, though, I just don't see how these old well-understood ideas from computer science are going to result in a new fundamental theory of physics.

Hopefully OP will response. I read ojna's post to mean that he prefers F# over Python, but that the rest of the job around the F# programming was miserable (maybe biz, org, process, and personality stuff was bad).