5 ms·

The article has "stay on the cutting edge". Ha! The real leading edge is work that is at least "new, correct, and significant" and published in a peer-review

by graycat 11y ago

The article has "stay on the cutting edge". Ha! The real leading edge is work that is at least "new, correct, and significant" and published in a peer-reviewed journal of original research in information technology, yes, and also powerful and valuable in practice. So, put on resume that have published in original mathematical statistics for detecting zero-day problems, in artificial intelligence, and in optimization, and still get silence. With Silicon Valley, leading edge has no connection with peer-reviewed original research! So, do a startup. Web development? Taught myself -- the Web page code works fine. The difficulty was a lot of poorly written documentaiton. The software was fast, fun, and easy. "Leading edge"? Yup -- my startup is based on some original applied math I derived, complete with theorems and proofs. Likely and apparently I'm the only person in the world who understands that work. Definitely leading. The problem my work solves, a lot of people have regarded as important. So far, I have the only good solution. Some of the people trying to solve the problem say that the solution I have is impossible -- they didn't understand what the heck they were talking about. They couldn't come up with a counterexample why not or a theorem of why can. The people just were not very good. Problem is, really, leading is not what Silicon Valley is looking for. They wouldn't know the real stuff about leading if it slapped them in the face. My best published papers, nearly no one in Silicon Valley has the prerequisites to read, and likely no one outside of Stanford or Berkeley and only a few people there. Instead of leading, they want two years of experience in just what the heck low grade, routine, simple minded, poorly documented API or OSS they are struggling with today. Those people are being self-taught on their way to the leading edge and are floundering around and stand not to get to the leading edge. E.g., they keep saying Bayesian and power law but don't know the difference between the weak and strong laws of large numbers or about conditional expectation, non-parametric hypothesis tests, the Radon-Nikodym theorem, anything about martingales, the Neyman-Pearson lemma, etc., enough about conditioning even to understand the definition of a Markov process, a basic result such as E[E[Y|X]] = E[Y] the role of sigma algebras, or any such things. Early in my career, I started in computing and started to get self-taught in the often associated applied math. I made good progress in some things but otherwise floundered around with low quality materials, etc. Then I went for a good Ph.D. and got one. Now I can say, nearly no one in Silicon Valley startups is in any significant sense even near the leading edge and, instead, they are just floundering around. Back to my startup -- a few good days and I should be able to go live.

puredemo 11y ago
>Back to my startup -- a few good days and I should be able to go live. Read back on your posts a few pages, respectfully requesting to be put on your alpha list as well. My email address is on my profile here, for the time being.
- graycat 11y ago
  You're on the list. Thanks.
auvrw 11y ago
> the role of sigma algebras i.e. the definition of probability? yea, i expected ppl to be more interested in that, myself otoh, be careful about minimizing other skill sets. doesn't take much more than basic arithmetic on the math side to understand the linux kernel, but the skills needed to be productive with such an engineering challenge aren't any kind of joke.
- graycat 11y ago
  It's easy enough in elementary treatments of probability to say what is meant by random variables X and Y are independent. And in both theory and applications, independence is just crucial. Well, what about uncountably infinite collections A and B of random variables being independent? Also, random variables that are functions of independent random variables are independent, etc. The sigma algebra generated by a random variable, or a collection of such, is important to be able to work with. E.g., just in the Poisson process, that is, the vanilla arrival process, need to talk about the time to the next arrival being independent of all of the past of the process. > Linux kernel Difficult? Right. No doubt by now it's huge, and I have to believe has many patches. And I have to suspect that it's not well documented. And people that use it especially want it for high end work where they make modifications, and then the problems of complexity, documentation, testing, etc. get much worse. Difficult? Yes. Leading edge, not in the usual sense of knowledge. More broadly, tough to make progress in any sense if still have to spend much effort with the Linux kernel -- we're supposed to be past that by now. This may be wrong for some niche need, but people should try to keep their careers out of such niches. Linux kernel performance? Might want some math there!

by graycat 11y ago

puredemo 11y ago

>Back to my startup -- a few good days and I should be able to go live. Read back on your posts a few pages, respectfully requesting to be put on your alpha list as well. My email address is on my profile here, for the time being.

graycat 11y ago
You're on the list. Thanks.

auvrw 11y ago

> the role of sigma algebras i.e. the definition of probability? yea, i expected ppl to be more interested in that, myself otoh, be careful about minimizing other skill sets. doesn't take much more than basic arithmetic on the math side to understand the linux kernel, but the skills needed to be productive with such an engineering challenge aren't any kind of joke.

graycat 11y ago
It's easy enough in elementary treatments of probability to say what is meant by random variables X and Y are independent. And in both theory and applications, independence is just crucial. Well, what about uncountably infinite collections A and B of random variables being independent? Also, random variables that are functions of independent random variables are independent, etc. The sigma algebra generated by a random variable, or a collection of such, is important to be able to work with. E.g., just in the Poisson process, that is, the vanilla arrival process, need to talk about the time to the next arrival being independent of all of the past of the process. > Linux kernel Difficult? Right. No doubt by now it's huge, and I have to believe has many patches. And I have to suspect that it's not well documented. And people that use it especially want it for high end work where they make modifications, and then the problems of complexity, documentation, testing, etc. get much worse. Difficult? Yes. Leading edge, not in the usual sense of knowledge. More broadly, tough to make progress in any sense if still have to spend much effort with the Linux kernel -- we're supposed to be past that by now. This may be wrong for some niche need, but people should try to keep their careers out of such niches. Linux kernel performance? Might want some math there!