Data science is one of the few fields where a physics degree does have at least some comparative advantage. In any engineering field, a hiring manager with a choice between an engineering graduate and a physics graduate is always going to choose the engineering graduate because they have specific training in the field. Part of the myth propagated by physics departments is that learning to solve a bunch of general problems will somehow make you qualified to solve very specific problems.
You are missing the argument. It is the same one that RAND https://www.rand.org/pubs/conf_proceedings/CF194.html gave when commissioned by congress to determine whether their was a “STEM shortage.” If there was a STEM shortage, wages would rise. Since rising wages aren’t observed, the situation is that the employers complaining of a shortage don’t want to pay the market rate. Instead they want extra-market forces (political programs to push workers into STEM training) to lower the market rate for them.
The comparison isn't between working at a software startup and delivering food. It is between working at a software startup and a different software job. The other software job is almost always better for the reasons listed above.
Any of the regulations predate most digital communications so there are a bunch of obsolete rules specifying the maximum symbol rate for digital transmissions rather then restricting the bandwidth which is what one actual cares about for sharing the available frequency space. This restriction artificially precludes use of many modern transmission schemes and hamstrings experimentation.
Useful description from the documentation about what is new about this as amateur packet radio has a long history.
The NPR protocol is designed to transfer IP data over radio links, in a bidirectional way (single frequency).
This protocol is in the middle between old packet radio (AX.25) and HSMM-Hamnet with Wifi equipment.
This protocol is designed by a HAM for HAMs. The project is 100% open source : specification, software, PCB.
This solution is complementary to HSMM-Hamnet (which uses Wifi equipments), on lower frequencies 70cm). Radio links on 70cm band is much more robust to obstacles.
The data-rate available is also much smaller, but yet useable. We can achieve several hundreds of kbps. The protocol is designed for “point to multipoint” topology, with 1 central relay (called Master) and several clients around.
Except go (and all other well defined games) are the most trivial of domains. For real business problems, the difficulty is most often trying to determine what the problem is and how it relates to the rest of the world.
Yeah that whole section on suspicious comparison looks like false positives by someone who didn't even bother to look at the code or doesn't understand it. Right after there is also the flagging of strod(rounded string) = 0.0 which I'm pretty sure is correct. They are checking whether the rounded string is exactly zero. Automated tooling is really only useful if you actually know what you are doing and can properly triage the output. "Look I ran this tool and it spit out some things" isn't very helpful.
Well the common description of a data scientist as "better at stats than a programmer and better at programming than a statistician" can also be reframed as "worse at stats than a statistician and worse at programming than a programmer" in most cases. I think it's typically a rebranding of PhD in a science field realized that market for that educational track is very small so rebranded as "hottest job of the next millennium"
The tax evasion argument doesn't seem to follow. It's not the scarcity that you want hard to measure it's the value. Scarcity in modern art is always assured by one of a kind or limited runs. If it's photorealistic it's much easier to judge the value than an abstract composition. How much should a shark in formaldehyde be worth? If it's in a natural history museum, then the price of the shark, the tank and the formaldehyde and any labor to prepare it. If it's in an art museum, who knows? Whatever someone paid for it.
No he's not getting hung up on the term vector. He is saying that SIMD is fixed but wider width and that is not equivalent to arbitrary width. For example, SIMD shuffle type instructions have no arbitrary length equivalent.
Yes, the paper is general and covers the case for arbitrary bit size. 64-bit would require a 128-bit approximate inverse in general and the multiplication during computing the mod would extend that out another 64-bits. So would require three registers and two more multiplications on x64 I think. Socidont think it gains you anything in the general case. Bit for some divisors it might beat the traditional Granlund-Montgomery-Warren method when you could use a smaller approximate inverse.
I think you may have missed his point slightly. He isn't questioning whether one can identify important results after the fact. He is questioning whether one can identify important discoveries that haven't yet been made. In many fields of inquiry one can assemble a list of important open questions. The issue is whether if you keep that list and check back whether 1) any of those questions have been answered 2) whether the answer turned out to be important 3) whether other questions not on the list were answered and turned out to be more important. He is arguing that 1 and 3 are the more likely outcomes so that a question being on such a list is likely not a good indicator that it is a fruitful avenue of research. Some examples would be things like grand unified physics theories, the Riemann hypothesis, whether P=NP. Will any of those be solved in the next 10 years, unclear, will something not on the list but also important be solved, chances are highly likely.
Peak sprint performance is in one's twenties, peak marathon performance is on one's thirties. Source: Usain Bolt and Elioud Kipchoge. You can find tons of finish times with age for running events online to show how the degradation in performance falls slowly with age for marathon times. It's actually hard to find information on sprint times for older competitors because the participation rate is much lower. There is a significant body of research that indicates humans have significant advantages for endurance performance versus other mammals. Born to Run is the classic popular book about the subject and has references to the primary sources.
It turns out that the effect is typically exactly the opposite. Design and process are already coupled where a given process will have design rules that must be adhered to to achieve a successfully manufacturable design. Intel only has to support their own designs so can have very strict design rules. Fabs like TSMC have to be more lenient in what they allow from their customers so have looser design rules that result in a less optimized process to achieve the same yield.
Yeah that is the exact strategy employed by 61 year old Cliff Young linked above, slow pace and no sleep. The article notes that is the strategy everyone uses now in that race as well. Also he was not competitive the second year.
The go compiler does dead code elimination. I'm not familiar with TinyGo but from the readme one thing it does do differently is drop the whole scheduler if there are no blocking go routines.
No. I think that is the exact point he was making that quantum key distribution is orthogonal to post-quantum cryptography and that the consensus is that poat-quantum cryptography will actually be used while quantum key distribution is basically an academic exercise because it doesn't have practical utility for the problem it is attempting to solve.
Well there are the fourteenth and fifteenth amendments and the voting rights act of 1965 that do establish the right to vote. But, yes those have all been significantly weakened by the judiciary, including to allow permanent disenfranchisement of convicted felons, which seems explicitly against the text of the fifteenth amendment.
Yeah I'm really confused why working with 4GB+ datasets in the browser is a current need. Why does everything end up in web browsers these days? Is it only because developers aren't familiar with anything else?
There are actually plenty of things people want that you can sell but they won’t buy, for example because they can get it cheaper or free somewhere else.