Yes. This is another in a long series of articles that describe the absolute worst possible example of a lecture, declare "lectures" awful and/or dead, and make some exhortation about learning styles (which have zero scientific basis, except for passive/active).
Learning style is just the converse of teaching style: the teaching style that works best for you. There's lots of scientific work in this area, and whole journals devoted to it, within the field of communication studies (now largely focused on distance learning).
He seems to be talking about "Into to" courses, where maybe he's right, but fancy demos aren't so useful for advanced topics.
I find the best sort of lecture is a recording of the best prof I can find talking to an audience of students who ask a lot of questions. Sure, I may occasionally have some question that wasn't asked in the recording, but as long as the course also has a way to ask that question, it's ideal.
Recorded lectures are great because there's just no tension between making notes and paying attention. I can rewind as much as I need to. I spent a lot of time recently watching lectures on quantum mechanics from Stanford's YouTube channel. Remarkably accessible. The ability to stop the lecture and work the math until I get it changes everything (math is the only useful language for understanding QM, but with dense notation it's very easy to get left behind).
What you're missing is: no one gives a shit unless they have some compassion for the ones you claim are oppressed in some area. People who are themselves actually oppressed sometimes find that compassion hard to come by.
Problem is: in Seattle, Amazon is paying way more than a living wage and even the median wage; both have sharply risen due in part to Amazon. This has been one cause of the sharp increase in homelessness in Seattle (the others being drugs and mental illness).
So, your theory is: a company paying its employees more causes poverty. Really? That's what you're going with?
VOIP needs different QOS even with minimal contention. Control plane stuff needs different QOS. Different QOS for streaming video can improve everyone's experience. It's not just about oversold networks - networks simply aren't perfectly reliable, and delivering packets late (vs dropping) may be good or bad, depending.
. Each individual customer may wish to apply QOS to their own data but since that would be customer controlled
Consumer networks should not/i> be built for the preferences of IT geeks, sorry. We get no special privilege at the expense of others.
Your lengthy explanation doesn't translate well into law, is my point. What about 0-rating? What about the next clever thing companies dream up to work around the regs. What about the regulatory change that the Cable company lobbyists get introduced.
I've never liked net neutrality. It's not a layer that the government needs to meddle at, because laws change too slow for technology, and big corporations are great at gaming laws. Yes, yes, the intention is great, but that's not how law works.
Networks need different QOS for different traffic - it's not like you can mandate treating all traffic identically, so you get into fine wording about "we mean you can't charge more to carry Netflix traffic, you know, like that". But whatever you write will have loopholes, and if it doesn't at first, give the cable lobbyists 5 years and it will.
This is an economic problem that won't have a technological fix. There are good ways to fix it, but federal regulations about how router firmware works isn't one of those ways. Just make the last mile a public utility already.
That's just another good reason why the network - the data center, the cables in the road etc. should be a public service like water pipes and electricity.
Not the datacenter. Just a termination facility for the last mile that any ISP can hook into. The last mile, specifically, is what needs to be a public utility. That's where the natural monopoly is. The rest the market really could sort out, as the barrier to entry would be small.
Not remotely credible. It presumes that a government agency would actually read comments from citizens. Not a feature of large governments. They are required to accept comments, of course. Of course.
Well, that's a bit of a programming style choice, at least for parameters. I've come to prefer different method names almost always instead of nullable parameters (the notable exception being backwards compatibility in formal interfaces) . It's only where I have some more abstract task system, when the parameters are data members in some class anyway, where nullable seems more sane.
Of course, for data members, that's all about whether null actually represents a valid state in the thing you're modelling.
I found that scoped objects were the right answer for the 80% case in the various domains I've worked in. I would love for "clean up when I exit scope" to be the default for object allocation, as it is in C++. But other choices should be declare-able.
The biggest miss in modern languages though is that "const, non-null" should be the default for all members and parameters. Give me an operator for those rare times when a reference is nullable ('?'), or a value is mutable ('!').
You're now talking about different problems. 1. Memory management. 2. Resource management.
Memory is just one resource. In some domains, it's not even the most frequently managed.
For resource management, there should be a language construct that brackets use of the resource.
The core concept here is "scoped objects". Most objects have a scope in which they are valid, and should be deterministically cleaned and freed on the way out of that scope (whether by return or throw). In C# and Java you can do this awkwardly with using/try-with-resources. In C++ you can do this effortlessly by making your primitives (pointers, file handles, etc) scoped objects.
Why not use the resource management idiom to manage memory? Because you don't necessarily release memory in the same place as you allocate it.
No different from any other resource. Some objects aren't scoped, and have lifetimes determined in some complex way. It would be nice to have reference-counting and GC as choices for explicit declaration for those times. C++ does refcount well, and that's the 90% case, but optional GC would sure be nice as another tool in the toolbox.
GC is certainly handy from time to time, but scoped objects are the 80%+ case in most domains.
One of the goals of the problem statement was to improve human productivity. Therefore GC would be a requirement. Not having to manually manage memory removes a large burden from programmers.
GC and manual alloc/free are not the only choices. GC only really takes care of memory anyhow - there are many sorts of open/close pairs besides memory: file handles, locks, etc. Solve the bigger pattern, not just specifically memory.
There are other, better ways to have the system deallocate things for you. GC is non-deterministic, which isn't great even if your system doesn't mysteriously pause every so often.
Video fingerprinting techniques are pretty good these days. If all they're doing is hash matching, they're not really trying.
Most storage service providers don't want people streaming video of any sort out of their public folders. Simple techniques that encourage people to put their pirate booty in a zip file or something at least stops the streaming.
Amazon offers the same. Managed, and HA if you want it.
And it's not Oracle - no contracts. Plus if you wisely hate Oracle, you have useful choices for other DBs. (Though if you want MySQL-compatible, Aurora is probably a better plan - hopefully Postgres will go live soon).
Sound like idealized history to me. If you're lazing about for significant periods, it meant you had more food than you could eat and all your tools and shelter were in good shape with ready spares (OK, we know of a couple of cultures where the men did lie about while the women worked continuously, but that''s anomalous).
I think it much more likely that people had their hands busy making tools, or using them, while sitting around chatting. Making anything with stone-age tools (including the tools) is quite labor intensive and time consuming, and nothing lasts long when used.
But maybe that's not what you meant by "serious" labor?
Either way though, speech recognition is much less useful without NLP, and I strongly suspect they have that NLP part, so it's kinda odd to not expose it.
I wonder whether they really do, in the way that Google.MS does. Google and MS have vast troves of search history to train machine learning algorithms against and build graphs of how language is used in practice. Amazon (nor Apple) doesn't have that data, so they'd have last-century-style hard-coded NLP.
Still, you'd think they could do better than what TFS describes - at least use a thesaurus.
That hasn't happened with Hollywood and Silicon Valley.
"Hollywood" films aren't really made in Hollywood any more. Canada and New Zealand are common.
What what I can see out my office window, half of Silicon Valley has already moved to Seattle. Tech in Austin is growing fast, but nothing like the new skyscraper every couple months you see in Seattle.
because realistically who the fuck in California wants to move to Texas?
About 60,000 people a year (about 30k net)? California has a net loss of about 100,000 people from domestic migration, while Texas has a net gain of about 100,000 people from same.
Using freedom of speech to try to shut down critics is both a hypocritical and losing argument.
Agreed. That's why I like Cloudflare for preventing "hate speech" sites, torrent sites, and so on from being shut down. Some of the site may be run by total assholes, but Cloudflare is doing good work.
Slashdot Mirror
And the result is: life gets better for everyone. That's what happens when successful companies pay their employees more. Life is not a zero-sum game.
Yes. This is another in a long series of articles that describe the absolute worst possible example of a lecture, declare "lectures" awful and/or dead, and make some exhortation about learning styles (which have zero scientific basis, except for passive/active).
Learning style is just the converse of teaching style: the teaching style that works best for you. There's lots of scientific work in this area, and whole journals devoted to it, within the field of communication studies (now largely focused on distance learning).
He seems to be talking about "Into to" courses, where maybe he's right, but fancy demos aren't so useful for advanced topics.
I find the best sort of lecture is a recording of the best prof I can find talking to an audience of students who ask a lot of questions. Sure, I may occasionally have some question that wasn't asked in the recording, but as long as the course also has a way to ask that question, it's ideal.
Recorded lectures are great because there's just no tension between making notes and paying attention. I can rewind as much as I need to. I spent a lot of time recently watching lectures on quantum mechanics from Stanford's YouTube channel. Remarkably accessible. The ability to stop the lecture and work the math until I get it changes everything (math is the only useful language for understanding QM, but with dense notation it's very easy to get left behind).
Or, you know, you commute from farther away, like the rest of us, and look for a job close to where you now have to live.
What you're missing is: no one gives a shit unless they have some compassion for the ones you claim are oppressed in some area. People who are themselves actually oppressed sometimes find that compassion hard to come by.
Problem is: in Seattle, Amazon is paying way more than a living wage and even the median wage; both have sharply risen due in part to Amazon. This has been one cause of the sharp increase in homelessness in Seattle (the others being drugs and mental illness).
So, your theory is: a company paying its employees more causes poverty. Really? That's what you're going with?
The shelter exists today, it's just in a temporary location (a hotel building that will eventually be demolished to build an Amazon building).
VOIP needs different QOS even with minimal contention. Control plane stuff needs different QOS. Different QOS for streaming video can improve everyone's experience. It's not just about oversold networks - networks simply aren't perfectly reliable, and delivering packets late (vs dropping) may be good or bad, depending.
. Each individual customer may wish to apply QOS to their own data but since that would be customer controlled
Consumer networks should not/i> be built for the preferences of IT geeks, sorry. We get no special privilege at the expense of others.
Your lengthy explanation doesn't translate well into law, is my point. What about 0-rating? What about the next clever thing companies dream up to work around the regs. What about the regulatory change that the Cable company lobbyists get introduced.
I've never liked net neutrality. It's not a layer that the government needs to meddle at, because laws change too slow for technology, and big corporations are great at gaming laws. Yes, yes, the intention is great, but that's not how law works.
Networks need different QOS for different traffic - it's not like you can mandate treating all traffic identically, so you get into fine wording about "we mean you can't charge more to carry Netflix traffic, you know, like that". But whatever you write will have loopholes, and if it doesn't at first, give the cable lobbyists 5 years and it will.
This is an economic problem that won't have a technological fix. There are good ways to fix it, but federal regulations about how router firmware works isn't one of those ways. Just make the last mile a public utility already.
That's just another good reason why the network - the data center, the cables in the road etc. should be a public service like water pipes and electricity.
Not the datacenter. Just a termination facility for the last mile that any ISP can hook into. The last mile, specifically, is what needs to be a public utility. That's where the natural monopoly is. The rest the market really could sort out, as the barrier to entry would be small.
Garbage Haulers of Milpitas?
Not remotely credible. It presumes that a government agency would actually read comments from citizens. Not a feature of large governments. They are required to accept comments, of course. Of course.
which turns out not to be that rare in practice
Well, that's a bit of a programming style choice, at least for parameters. I've come to prefer different method names almost always instead of nullable parameters (the notable exception being backwards compatibility in formal interfaces) . It's only where I have some more abstract task system, when the parameters are data members in some class anyway, where nullable seems more sane.
Of course, for data members, that's all about whether null actually represents a valid state in the thing you're modelling.
I found that scoped objects were the right answer for the 80% case in the various domains I've worked in. I would love for "clean up when I exit scope" to be the default for object allocation, as it is in C++. But other choices should be declare-able.
The biggest miss in modern languages though is that "const, non-null" should be the default for all members and parameters. Give me an operator for those rare times when a reference is nullable ('?'), or a value is mutable ('!').
You're now talking about different problems. 1. Memory management. 2. Resource management.
Memory is just one resource. In some domains, it's not even the most frequently managed.
For resource management, there should be a language construct that brackets use of the resource.
The core concept here is "scoped objects". Most objects have a scope in which they are valid, and should be deterministically cleaned and freed on the way out of that scope (whether by return or throw). In C# and Java you can do this awkwardly with using/try-with-resources. In C++ you can do this effortlessly by making your primitives (pointers, file handles, etc) scoped objects.
Why not use the resource management idiom to manage memory? Because you don't necessarily release memory in the same place as you allocate it.
No different from any other resource. Some objects aren't scoped, and have lifetimes determined in some complex way. It would be nice to have reference-counting and GC as choices for explicit declaration for those times. C++ does refcount well, and that's the 90% case, but optional GC would sure be nice as another tool in the toolbox.
GC is certainly handy from time to time, but scoped objects are the 80%+ case in most domains.
One of the goals of the problem statement was to improve human productivity. Therefore GC would be a requirement. Not having to manually manage memory removes a large burden from programmers.
GC and manual alloc/free are not the only choices. GC only really takes care of memory anyhow - there are many sorts of open/close pairs besides memory: file handles, locks, etc. Solve the bigger pattern, not just specifically memory.
There are other, better ways to have the system deallocate things for you. GC is non-deterministic, which isn't great even if your system doesn't mysteriously pause every so often.
Drawing more attention to a meme isn't really the best way to kill it off. Now it's all Jesus-motif Pepes.
It's not easy being green.
Video fingerprinting techniques are pretty good these days. If all they're doing is hash matching, they're not really trying.
Most storage service providers don't want people streaming video of any sort out of their public folders. Simple techniques that encourage people to put their pirate booty in a zip file or something at least stops the streaming.
Amazon offers the same. Managed, and HA if you want it.
And it's not Oracle - no contracts. Plus if you wisely hate Oracle, you have useful choices for other DBs. (Though if you want MySQL-compatible, Aurora is probably a better plan - hopefully Postgres will go live soon).
Sound like idealized history to me. If you're lazing about for significant periods, it meant you had more food than you could eat and all your tools and shelter were in good shape with ready spares (OK, we know of a couple of cultures where the men did lie about while the women worked continuously, but that''s anomalous).
I think it much more likely that people had their hands busy making tools, or using them, while sitting around chatting. Making anything with stone-age tools (including the tools) is quite labor intensive and time consuming, and nothing lasts long when used.
But maybe that's not what you meant by "serious" labor?
Either way though, speech recognition is much less useful without NLP, and I strongly suspect they have that NLP part, so it's kinda odd to not expose it.
I wonder whether they really do, in the way that Google.MS does. Google and MS have vast troves of search history to train machine learning algorithms against and build graphs of how language is used in practice. Amazon (nor Apple) doesn't have that data, so they'd have last-century-style hard-coded NLP.
Still, you'd think they could do better than what TFS describes - at least use a thesaurus.
That hasn't happened with Hollywood and Silicon Valley.
"Hollywood" films aren't really made in Hollywood any more. Canada and New Zealand are common.
What what I can see out my office window, half of Silicon Valley has already moved to Seattle. Tech in Austin is growing fast, but nothing like the new skyscraper every couple months you see in Seattle.
because realistically who the fuck in California wants to move to Texas?
About 60,000 people a year (about 30k net)? California has a net loss of about 100,000 people from domestic migration, while Texas has a net gain of about 100,000 people from same.
Using freedom of speech to try to shut down critics is both a hypocritical and losing argument.
Agreed. That's why I like Cloudflare for preventing "hate speech" sites, torrent sites, and so on from being shut down. Some of the site may be run by total assholes, but Cloudflare is doing good work.