HNHacker News
TopNewBestAskShowJobs

NY_USA_Hacker

369 karma · joined January 21, 2011

xyz
submissionscomments
NY_USA_Hacker··on Ideas Matter
"This is rather an attack on a strawman. No one thinks that ideas don't matter at all."

Not really: The contempt for ideas among venture partners is overwhelming. E.g., KPCB VC Doerr at

http://www.youtube.com/v/nBvuirDPHKA&hl=en_US

claimed that "Ideas are easy" and "plentiful" and "Execution is everything.". Union Square VC Wilson at

http://www.avc.com/a_vc/2011/04/how-to-allocate-founder-and-...

said "Ideas are pretty much worthless.".

For a diplomatic view of what Doerr and Wilson said, mostly by an 'idea' they mean some nearly instant guess at a one sentence description of what would make a good description of a product or service to a customer, and the hope is that the business would be good.

Of course, still these venture partners are being silly: The product or service selected to sell to customers is crucial. Broadly, many products and services can be offered that essentially no customers would want even for free. And there are products and services that many customers would want, but the price would have to be too high. And there can be other problems with the product or service envisioned.

Then there's the claim that execution is so difficult: Obviously this is nearly total nonsense as anyone in the US can easily see. E.g., the US is just awash in businesses, millions of them, supporting families, coast to coast, on Main Street in villages, towns, cities. These businesses sell carryout pizza, repair roofs, mow grass, pour concrete, install or repair plumbing or HVAC, maintain cars, do auto body repair, buy products at a low price per unit in large quantities and sell them at a higher price per unit in smaller quantities (per sale), and on and on. E.g., look at the businesses on the TV program 'Diners, Drive-ins, and Dives' and identify the 'exceptional' abilities in "execution" they have mastered?

From living in several major parts of the US, I have yet to see such a US Main Street where the rate of business turnovers was high or where most new businesses start and fail quickly. So the failure rate has to be fairly low. If execution were so difficult, then the failure rate of these Main Street businesses would be much higher.

Why venture partners are so consistent in getting these issues of Business 101 wrong is curious, frustrating, and disconcerting.

Here is a much better description of what is true: Bad ideas for products and services are easy, and then execution is difficult. Good ideas for new products and services are difficult, and then execution is routine. As Main Street shows, the US is just awash in the ability to do routine execution well.

Then also Kumar misses a crucial point for a business based on a new product or service. So, he writes:

"In fact, most founders tend to take it to the other extreme that they don’t want to share their idea (subject of another future blog post on why Stealth is Overrated)."

So all he sees for an 'idea' is just the description of the product or service. Wildly wrong.

There is a much older, and much more appropriate, description of how to start a successful business with something new:

(1) Problem Idea.

Pick a good problem to solve. To have a really successful business, the problem has to be 'big' enough, e.g., with a few customers ready to pay a lot or a lot of customers willing to pay at least a little.

IBM picked using punched cards from Hollerith to replace huge rooms of clerks doing routine business record keeping.

Xerox picked office photocopying.

Ford picked a car for everyman for a few hundred dollars.

DEC picked small computers for scientists and engineers.

FedEx picked overnight delivery of small packages with excellent package tracking.

Wal-Mart picked big box general store retailing with excellent buying and logistics and locations in smaller communities.

Google picked Internet search based on keywords much like an old library card catalog subject index but aided with 'page rank' as a measure of popularity.

(2) Newness Idea.

Main Street businesses often do well enough without much attention to newness because they can have a 'geographical barrier to entry' and, thus, no competition more than 50 miles away. So do well against competition in a radius of 50 miles and do okay.

Without a geographical barrier to entry, to help avoid competition, it is important to have something new about the business that helps revenue and/or earnings, and this something new should be difficult to duplicate or equal.

For IBM, its punched cards were new. Then IBM did exceptionally well with product quality and customer service.

For Xerox, the electromechanical internals of their machines were new.

For FedEx, they used the CAB regulation for unscheduled air taxi to be the only nationwide company both to operate trucks and fly airplanes and, thus, have full control over the packages from origin to destination and excellent ability to track the packages.

DEC did well with the new technology of printed circuits with discrete components.

(2) Solution Idea.

One source of something new is the solution, that is, the means of supplying the product or service.

If the problem to be solved is really big but not yet well solved, then likely plenty of people have seen the problem but have not been able to find an effective solution.

In now very old terms, the solution might be something that deserves a patent (Xerox) or that should be protected as a trade secret (Google). The solution might be the main source of newness for the business.

Here is the big, HUGE mistake of information technology venture capital: The assumption is that the core technology is just routine software with nothing new that is difficult to duplicate or equal. That is, the technology has no significant 'ideas'.

Instead, the core technology might have solid, new, powerful, valuable 'ideas'. Various parts of our society -- DoD, NSF, NIH, biotech venture capital, the patent system, etc. -- recognize the importance of such ideas, but information technology venture capital, with outrageous hostility, rejects any such possibilities.

The nonsense is too uniform to be just 'natural'. Instead it appears that some LPs had lunch and decided to take a preemptive negotiating position to deprecate to worthlessness anything new from entrepreneurs in information technology.

Here's some of where such venture capital is making a mistake:

(1) New Ideas.

The US research universities remain crown jewels of civilization, and the leading graduate programs provide excellent starts in how to find powerful, valuable new ideas. Indeed, the common criterion for a Ph.D. is some original research worthy of publication, and the usual criteria for publication are new, correct, and significant.

The power of such research has been well appreciated by the US DoD for about 70 years and by the US NIH for nearly that long.

For US information technology venture capital to ignore or deprecate the power and value of such research will have to prove to be a long walk on a short pier.

(2) Solving a Big Problem.

Such new ideas can be the crucial, core 'secret sauce' that is difficult to duplicate or equal and that permits solving a big problem.

(3) Starting a Web Based Business.

Now it costs less in capital equipment to start a new Web based business than a carry out pizza shop, grass mowing service, auto repair shop, auto body shop, etc. These old Main Street businesses essentially never get equity funding, and it's getting to be a strain just why new Web based businesses should seek equity funding.

(4) Execution.

For a new Web based business with a few servers in a spare room, small commercial space, or colocation site, the challenge of execution stands to be less than for most Main Street businesses. Or, if Main Street can do the execution, then a researcher with a powerful, valuable new idea and a Web based business should be able to also.

So, for the entrepreneur, the opportunity is:

(1) Pick a big problem not solved well.

(2) Do some original research and find better means of solving that problem. The research is the main 'idea' and should be difficult to duplicate or equal and protected as a trade secret. That all such ideas are "easy", "plentiful", and "worthless" is insulting, brain-dead, absurd, and outrageous. Nor will such ideas be found by 'pivoting' during 'execution'.

(3) Implement the research in software and run it on servers in a Web based business.

(4) Bring up the Web site, go live, get users, run ads, get ad revenue, and have a business.

For a Series A round, those venture partners will want the business to have so much 'traction' that is already is or soon will be throwing off cash enough to let the entrepreneur grow the business quickly and have a nice living.

E.g., maybe early on the Web site sends 10 Web pages a second, 24 x 7, with three ads per page and with good enough ad targeting for $1 in ad revenue per 1000 ads displayed. Then the revenue in a 30 day month would be:

1 * 3 * 10 * 3600 * 24 * 30 / 1000 = 77,760

dollars. So, the entrepreneur is already doing as well as a Main Street entrepreneur with several carry out pizza shops.

For such an entrepreneur, just why would they want to report to a board of directors with venture partners who just ignore the potential of new ideas from research and generally insist that ideas are "easy", "plentiful", and "worthless" and that execution, done well enough by millions of US Main Street entrepreneurs, is "everything"?

Net, the flip side of this outrageous information technology venture partner brain-dead, deprecating, insulting nonsense is one heck of a collection of good business opportunities.

This incongruous situation is close to inevitable: Really sweet business opportunities can be there only if they are not easy to see. Apparently it is overwhelmingly attractive just to wallow in the nonsense that ideas are easy and execution everything.

NY_USA_Hacker··on [dead]
Same as always: All three tuning forks are close together in the same lab in an inertial frame, etc. You see the same ratios as the lab does. If you compare with some tuning forks on your space ship, then just use special relativity to compare the ticks on your space ship versus the ones in the lab. If the tuning forks are 'standard', say, some spectral lines, the you can see your time dialation.
NY_USA_Hacker··on [dead]
My old analogy:

Start with a barter society. Suppose one apple is worth two oranges, and one orange is worth three lemons. Then one apple is worth six lemons. So, say that one lemon is worth one penny. Then one orange is worth three pennies, and one apple is worth six pennies. So, we now have prices for everything.

So, what is one penny? It's a 'common ratio'.

My analogy is the same for time: One tuning fork vibrates at 100 Hz, another at 200 Hz, and another at 600 Hz. So, call one time 'tick' the time of one period of the 600 Hz tuning fork. Then the one period of the 200 Hz tuning fork is three ticks, and, of the 100 Hz tuning fork, six ticks. So, one 'tick' of 'time' is a common ratio.

Seems okay to me.

NY_USA_Hacker··on Author of LuaJIT explains why compilers can't beat hand-coded assembly
It's a very old story: Hand coded assembler can commonly get a factor of several faster execution time on compiled code and more when using special instructions on special data. But on large programs, no, because some of what compilers and runtime libraries do for large programs is too difficult for hand coded assembler. Indeed, some of what compilers do in register allocation (and maybe cache usage) will be too difficult for hand coded assembler.

More generally, now the execution time of a small, single threaded program is nearly never of much concern. E.g., when have a processor with dozens of cores, it's easy to let a small, single threaded program have a core and then f'get about it.

Execution time remains important but now mostly just for large, multi-threaded programs.

NY_USA_Hacker··on David Einhorn Loves Microsoft, Hates Steve Ballmer
Einhorn should be careful what he wishes for: Microsoft is an enormously profitable company, and Ballmer is raking in the money. For this, he has at least decent understanding of that business. It would be tough to find a replacement, other than Gates, who would understand the present business nearly as well as Ballmer. Someone who didn't understand the business could seriously hurt it.

If Einhorn wants Ballmer to do better on some projects, then the result could be one or a few new VPs and projects within Microsoft.

Net Einhorn should be talking about maybe some new VPs, not a new CEO.

Actually, likely Einhorn doesn't know even as much as dip squat about the astoundingly successful parts of Microsoft's business. People who do understand Microsoft's business might want to see Gates return or some new VPs but will be very, VERY careful about getting rid of Ballmer.

NY_USA_Hacker··on The pitfalls of jargon and the damage it does to mathematics
I've had no interest in the complex half, but I can claim to have done well enough with the real half. I like it.

He does especially well with Fourier things, which is not surprising considering some of his other work,

He gives the von Neumann proof of the Radon-Nikodym theorem. I prefer the approaches with more steps with each step smaller, but it's great to see the von Neumann masterstroke.

Rudin's chapter on Banach space knocks off the main results quickly and cleanly.

His construction of Lebesgue measure early on is a bit long but surprising and, net, shorter than the other ways to do it.

Curiously he has a nice chapter with more general results on change of variable in multiple integration.

He does well with the various duality results, which are surprising and important.

A good competitor, of course, is Royden, 'Real Analysis'.

NY_USA_Hacker··on The pitfalls of jargon and the damage it does to mathematics
If you want something written as well as Halmos, then start writing, and good luck!

"Historical accident": Well, sure, in part nearly all the pure math departments pushed out any such topics!

You are correct: Optimal control has been mostly in advanced parts of electrical engineering.

So, optimal control in EE was an example doing math outside math departments. Of math done outside math departments, control theory is relatively good mathematically.

And, yes, there should be applications elsewhere.

For physics, yes, the deterministic theory of optimal control has been seen as replacing the older calculus of variations which goes back to Newton.

My references are old.

More recent work on stochastic optimal control has been by R. T. Rockafellar at University of Washington.

For

Stuart E. Dreyfus and Averill M. Law, 'The Art and Theory of Dynamic Programming', ISBN 0-12-221860-4, Academic Press, New York.

'dynamic programming' is a big part of the discrete time versions of optimal control. It can be stochastic or deterministic. The linear-quadratic-Gaussian (linear 'plant' or system, quadratic cost to be minimized, and Gaussian exogenous random variables) has 'deterministic equivalence' -- nice -- and this book treats it. The book is a good, elementary start. Apparently Dreyfus was a R. Bellman student.

For

L. S. Pontryagin, V. G. Boltyanskii, R. V. Gamkrelidze, and E. F. Mischenko, 'The Mathematical Theory of Optimal Processes', ISBN 0-470-69381-9, Interscience Publishers, John Wiley & Sons, New York.

it was, of course, the main source of the Pontryagin maximum principle, and, thus, a swift kick in the back side for parts of US aerospace in the 1960s.

For

Michael Athans and Peter L. Falb, 'Optimal Control: An Introduction to the Theory and Its Applications', McGraw-Hill Book Company, New York.

Athans was long in EE at MIT and did some military work, e.g., on parts of the C5A airplane. Falb was at Brown's Division of Applied Mathematics.

One Athans story was the 'control' for least time to climb to, say, 100,000 feet for an F-4: Go up to a few thousand feet, go into a dive, get supersonic, get the lower drag of a few hundred knots above Mach 1, and then with the lower drag continue supersonic to the final altitude. I was not able to know if that 'control' idea was just intuitive or directly from computation and the Pontryagin maximum principle which is, after all, just a necessary condition, i.e., local optimality.

Also in that division at Brown, see the papers of Harold Kushner. As I recall, he wrote out a stochastic version of the Pontryagin maximum principle.

For

E. B. Dynkin and A. A. Yushkevich, 'Controlled Markov Processes', ISBN 0-387-90387-9, Springer-Verlag, Berlin.

there are connections with economic planning.

For

Dimitri P. Bertsekas and Steven E. Shreve, 'Stochastic Optimal Control: The Discrete Time Case', ISBN 0-12-093260-1, Academic Press, New York.

the math is done carefully. So there is a lot of attention to measurability. Part of the reason is the issue of 'measurable selection': This can be a deep subject, but often a relatively simple way out is via regular conditional probabilities as in

Leo Breiman, 'Probability', ISBN 0-89871-296-3, SIAM, Philadelphia.

For

David G. Luenberger, 'Optimization by Vector Space Methods', John Wiley and Sons, Inc., New York.

this has likely the easiest mathematical treatment of deterministic optimal control and also Kalman filtering and also the math needed for 'least action' in physics.

For

E. B. Lee and L. Markus, 'Foundations of Optimal Control Theory', ISBN 0471-52263-5, John Wiley & Sons, New York.

this tried to be clean mathematically when it was written. At the beginning, should know some relatively advanced results in ordinary differential equations, e.g., as in

Earl A. Coddington and Norman Levinson, 'Theory of Ordinary Differential Equations', McGraw-Hill, New York.

For more, you can consider non-linear filtering and connections with mathematical finance.

The above is just from my bookshelf. Likely now a better bibliography could be assembled via the Internet which actually is at least a little larger than my bookshelf!

NY_USA_Hacker··on The pitfalls of jargon and the damage it does to mathematics
Errata: Of course, in

"In particular, in Baby Rudin, if a real valued function of a real variable is continuous on a closed set, then the function is also uniformly continuous, is bounded, actually achieves its upper and lower bounds, and its Riemann integral exists and is finite."

replace

"on a closed set"

with

"on a closed and bounded set".

NY_USA_Hacker··on The pitfalls of jargon and the damage it does to mathematics
In an overwhelmingly important sense, pure math has the least 'jargon' of any field. The reason is, math exposition has an absolute requirement: Each use of a word not the same as just its dictionary meaning becomes a 'term' and just MUST be defined, and 'well-defined', before it is used. So, what the article is calling 'jargon' is just such terms, but in well written pure math they all have rock solid definitions, and math is the unique field where this precision is true.

For the rest of the complaints about 'jargon', these are essentially just necessary given that math, due to its precision and age, is now by far the deepest field of study. So, there are a LOT of terms defined.

If the article wanted to claim that for each math result or paper there is a 10,000 foot intuitive view that can be explained in 90 seconds to just anyone on a street, then okay, but the article did not do this. This claim is not quite true, but it is close enough so that some such explanations, maybe with some pictures, and some connections with more elementary topics, can be useful, even in research papers. But the article didn't make this claim, either.

NY_USA_Hacker··on The pitfalls of jargon and the damage it does to mathematics
"Any HN'ers tried to hack their way through Baby Rudin?"

Yup. Been there; done that!

Yes, it needs some 'explanation':

Q. Why do we care about metric spaces?

A. Eventually they are important, e.g., in Hilbert and Banach spaces. For now, they are the vanilla case of a topological space.

Q. Why do we care about topological spaces?

A. Because they are the nicest way to describe continuity. So a function from one topological space to another is 'continuous' iff (from Halmos: if and only if!) for each open set in the range space its inverse image under the function is an open set in the domain space. That's a nicer, and more general, definition than converging sequences (which are not quite as powerful, but examples are a bit bizarre) or epsilon-delta arguments from limits in calculus.

Q. An 'open' set?

A. A 'topology' is defined as a collection of sets it calls 'open'. So, really the topology is defined just by some open sets. A good and easy example of an open set is an open interval, say, (0,1) on the real line. That set is 'open' in the 'usual' topology (in principle there can be others) for the real line.

Q. Why do we care about continuity?

A. As in G. Simmons, the two main pillars of the 'analysis' part of math are linearity and continuity. They are assumptions that commonly hold in practice and that give some of the most powerful and useful results. In particular, in Baby Rudin, if a real valued function of a real variable is continuous on a closed set, then the function is also uniformly continuous, is bounded, actually achieves its upper and lower bounds, and its Riemann integral exists and is finite. All nice stuff.

Continuing, all such functions form a vector space (a 'linear' space); the difference to two such functions is again such a function; and the maximum value of the difference forms a norm. So, those functions form a normed linear space. If a sequence of such functions converges in this norm, then the limit is also such a function -- i.e., a uniform limit of continuous functions is continuous. So, the space is 'complete' in this norm. So, we have the leading example of a Banach space (a complete, normed linear space). Then we can show the Hahn-Banach result and see the tasty, ice cream and cake dessert examples in Luenberger, 'Optimization by Vector Space Methods' including Kalman filtering, results for 'least action' in physics, more in deterministic optimal control, etc. So, these are some of the coming attractions of those first chapters in Baby Rudin on metric spaces, topology, open sets, continuity, etc.

Q. A 'closed' set?

A. Back to the real line where (0,1) is an open set, [0,1] is a closed set, So, it's closed and bounded. Presto: It is the leading example of a 'compact' set, which is almost as nice to work with as a finite set. In particular, any infinite subset of a compact set has a limit point in the set which is the key to uniform continuity, the existence of the Riemann integral, and more. The nicest definition of a compact set is that each open cover has a finite subcover, and this definition works also in metric spaces and topological spaces.

So here is an explanation of what the first chapters of Baby Rudin are all about. Print a copy and pin it to the wall in any math department that teaches a Baby Rudin course!

NY_USA_Hacker··on The pitfalls of jargon and the damage it does to mathematics
Well, Halmos wrote the first version of 'Finite Dimensional Vector Spaces' while he was working as an assistant to von Neumann at the Institute for Advanced Study. So the book is doing finite dimensional linear algebra over the real and complex numbers using the techniques of Hilbert space theory, a von Neumann speciality. Once von Neumann had to explain what he meant by 'Hilbert space' to Hilbert!

But that Halmos was a good writer is no joke: He was one of the best of the 20th century.

NY_USA_Hacker··on When Will I Use Math?
The 'job' qualification that worked for me was 'computing' the time in two weeks I went on seven interviews and got five offers.

Since then, with applied math, a Ph.D., etc. getting a job has been a scavenger hunt.

A good background in stochastic processes was supposed to be good for a career on Wall Street, but Goldman Sachs and Morgan Stanley didn't agree.

I'm no longer looking for a job. Instead I'm starting a business and now am quite far along. I may have a nice stream of revenue later this year.

Yes, for this thread, a background in parts of control theory might help in work in robotics. But some of the other considerations remain: Will the hiring manager be comfortable with a subordinate who knows more? Will the HR people recognize good qualifications when they see them? Is it really good to recommend to a young person to study control theory on some possibility that it will help getting a job in robotics?

There is a bigger issue that I omitted: Math in the US is still dominated by 'pure math' (some of which is crucial) which as a field is still mostly interested in "the analytic algebraic topology of the locally Euclidean metrization of infinitely dfferentiable Riemannian manifolds" (an old joke, extra credit for knowing the source), etc. and just HATES any suggestion of 'professional' education. Moreover those academic departments don't have even a weak little hollow hint of a tiny clue about applications outside of pure math or math physics. The pure math departments worked hard to push out probability, statistics, stochastic processes, optimization, control theory, continuum mechanics, numerical methods, graph theory, etc. Yes, there have been some exceptions at Brown and Courant. Yes, for some years there were efforts in 'the mathematical sciences'.

So, in particular, in business, they will take seriously professional work by lawyers and accountants but not mathematicians (main exception, actuaries in insurance). Part of the difference is that lawyers have a rule that a lawyer working as a lawyer can be supervised only by another lawyer. Net, math just has no 'professional' status in business, and that is a big problem for math helping a career in business.

Eventually there was the David Report that claimed that pure math just was not well enough connected with applications to be very relevant, and the NSF then started cutting back on grants for US pure math.

Science, engineering, computer science, statistics in the social sciences, etc. have had to do their own math and generally have done it poorly. It's a bad situation.

Net, for this thread, for a person planning their career, the video was not good; for all the potential math has, a person who pursues math in business will be mostly out on their own; the math will be in some secret sauce that neither the customers nor the venture partners know about.

NY_USA_Hacker··on When Will I Use Math?
Some of the qualifications of some of the YC partners are exceptional for venture partners: You might be correct.

But, to add a little detail, it appears that information technology (biotech might be significantly different) venture funding has been reduced to a large font on one side of a 3 x 5" index card: For a seed round, see a demo of the software. For a Series A, see 'traction', that is users and revenue both growing quickly. For a Series B, see good accounting numbers. For a Series C, are basically just buying part of an on-going company. Could teach a dog that 3 x 5" card in a weekend.

The Series A criteria are so uniform that my guess is that they are 'suggested' by some of the more important LPs.

But, for anyone with a serious background in math, science, or engineering, (information technology) venture funding is a shock because, if only as a special case of the 3 x 5" card, the partners will essentially never take seriously a technical review of the project on paper. This is a shock because essentially all serious work in math, science, and engineering is done just on paper, not with prototypes or 'market ready' products.

My joke about venture capital is that they would have said, "You build and test one and get one more ready for delivery, and we will buy half the gas for the Enola Gay".

But the bomb was a big waste of money? The bomb cost about $3 billion in 1940s money (see Richard Rhodes), and that's $3000 for each of the 1 million US soldiers that might have been killed or injured in an invasion of the home islands of Japan, that is, a grand bargain just as money.

So, (information technology) venture capital just won't judge based on looking in detail at the core 'secret sauce' intellectual property and, in particular, won't consider math.

So, if the math helps with the traction or accounting numbers, then fine. The entrepreneur can tell the equity funders that the project is just 'software' and otherwise keep what is crucial in the secret sauce just between his own ears. If he has an algorithm that shows P = NP (not what I have) and is blazingly fast on the full range of NP-complete problems, then don't tell anyone and just let the software run and do logistics, manufacturing scheduling, etc.

In my project, the math is the crucial work and the main promise of a valuable company quite beyond just some Series A 'traction' numbers. But the only person in business who can see this promise now is me, and this is at least a surprise.

In the end, for me, now, that the venture partners will ignore or just hate the math is no longer very relevant: I have the math and the corresponding software long since done, and it was fairly easy given my background. The rest is just routine programming. Since I'm using Windows, .NET, ASP.NET, and ADO.NET, I've had to learn those topics. This learning has been routine but slow. But a few more Web pages, and the software will be ready for at least first production. Then get and load some good initial 'base data', and I will be able to go live. If the project is good, then there will be good 'traction' numbers. Then I could raise a Series A, but if the project is good soon it should throw off enough cash that I won't t need a Series A. If I get a good business without equity funding, then the venture partners won't be part of it because they refused to evaluate the real promise, the math, and waited until the business was so far along it didn't need equity funding.

Then, just why would I want a Board that hated math?

Net, for this thread, for an information technology entrepreneur, the math just has to be something the entrepreneur keeps between his ears, and that is a piss poor career direction to recommend to young people.

NY_USA_Hacker··on When Will I Use Math?
Q. "When will I use math?"

A. In business, whenever you can find a use and start a company to pursue it.

Note:

(1) I know some math: My Ph.D. dissertation was on the applied math of stochastic optimal control. The work was advanced enough mathematically to address measurable selection, solved a real problem, and made some contributions to making such calculations faster.

(2) I'm pursuing a use of math in business. My current work is an information technology startup based on a Web site. The key to the promise of the work is some original math I did based on some advanced prerequisites.

So, I know something about applications of math.

Jobs? Early in my career, I knew some computing, was at GE which was shrinking (again) in computing, sent some resume copies, and in two weeks got seven interviews and five offers. The computing worked for a job.

Broadly the video is nonsense, contemptible, deceptive, misleading, badly supported nonsense. Students need to be warned.

Wall Street? In graduate school, Wall Street was one of the career directions I had in mind. So, I got a relatively good background in stochastic processes. Later I wrote Fisher Black (as in the Black-Scholes option pricing model) at Goldman Sachs and got a nice letter back saying that he saw no opportunities to apply math on Wall Street.

I interviewed in computing at Morgan Stanley, showed some nice applied math I'd done, and mentioned that I wanted to work on automatic trading but got no interest.

I got a call from Google, and they wanted only C++ programming.

I got a call from Microsoft and explained some work I'd done, and published, that would help Microsoft but never got past the first 'phone-screen' girl.

Early in my career, there was some interest in applied math, but this interest was essentially only for work paid for by the US DoD and for the Cold War. The US DoD has done a lot of good applied math, but those successes have had essentially zero influence in business.

Can look at the Web sites of 500+ US venture capital firms, nearly all in Silicon Valley, near Boston, or in New York City, look at their investment interests, and never see the words math or mathematics. Chris Sacca dropped out of a math Ph.D. program and may have the best background in math of any venture partner in the US. His biography may have the only mention of math on a venture firm Web site.

The interest of math in information technology venture capital is somewhere between zero and deeply negative. E.g., I got back from Ron Conway's son at Andreessen Horowitz

"your company is outside our area of expertise".

By "expertise" he had to mean the math, not the Web site.

Yes, in topics central to Hacker News, there should be interest in math for 'machine learning, data mining, big data, ad targeting' etc., but there is not. Instead, the world would rather stumble along with intuitive approaches to these topics instead of solid (powerful, valuable) math approaches.

Here's the main point: In business, the US is still organized mostly like the factory floor of 100 years ago where the supervisor knew more and the subordinate was there just to apply muscle to the ideas of the supervisor. And that 'inequality in technical knowledge' continues all the way to the CEO and COB who want to believe that they know the most.

So, no one in middle management wants to bet their budget or career record on some project using some math they don't understand; they don't get bad marks for such projects they do not pursue.

Essentially only a CEO could sponsor a project with some significant math, and CEOs rarely sponsor projects directly.

In venture funded information technology entrepreneurship, the 'technology' is supposed to be just routine software that could be understood by a bright, self-taught middle school hacker and, thus, can safely be ignored as crucial 'intellectual property' by any Ivy League history major, Harvard MBA venture partner.

Broadly, bringing high technical expertise as an employee to a business is a fool's errand. Yes, there are a few exceptions, e.g., some technical topic deep in the CIO's organization, but generally the employee is supposed to know less, not more.

An employee who, on the side or whatever, starts a good project based on math will find that their management doesn't like the project. If the employee proposes the project more broadly, then they will be told not to communicate outside of their management chain or face getting fired.

Once as a professor, hired in optimization, I wrote a research paper in optimization based on some math, with theorems and proofs. My department reviewed the paper and concluded that it was "not publishable". It published right away, without significant revision in the 'Journal of Optimization Theory and Applications'. At Yorktown Heights, we were doing some work in server farm and network monitoring of health and wellness using artificial intelligence. I found a much more powerful approach with some original work using some advanced topics in stochastic processes, and the lab concluded that the work was "not publishable". The work published right away without significant revision in 'Information Sciences'. The reason: The people I was working for didn't know math and didn't like it.

Really, if math is to have a significant role in a business, then the main mathematician has to be the founder, CEO of the business, e.g., Jim Simons, Andrew Viterbi. Otherwise, f'get about it.

Broadly, math is by far the most exacting and difficult academic subject, is rarely pursued by students beyond what is needed for K-12 teaching, physics, engineering, or basic statistics, and is just not known, liked, wanted, or even tolerated in society more broadly.

For a good career, mostly an electrician's license is a better foundation than a Ph.D. in math, electrical engineering, or anything else.

My view is that math is by a wide margin the most powerful and valuable topic for the future of information technology entrepreneurship for at least the first half of this century, but to make this promise real the mathematician has to be the founder, CEO of the company, and the math will not be an advantage in seeking equity funding.

To make this promise understood, about the best I can do is be successful financially, buy a yacht over 200' long, and give some interviews on the yacht cruising in Long Island Sound.

Broadly our society does not value, want, like, or tolerate math.

NY_USA_Hacker··on Don Dodge on The Next Big Thing: Four tests for if you have a good startup idea
Mostly nonsense.

Dodge is wallowing in waste, and from that 'context' what he is saying sounds okay.

But the truth is different:

First, the most important thing to have in business is good luck. Since we can do little or nothing to control that, let's set it aside in favor of things we can control.

Second is the 'idea'. If lots of other people can have the idea and execute it, then the business promises to have too much competition. Indeed, there may be established competitors now difficult to start against. So, setting luck aside, the 'idea' has to be one very few people can understand and execute.

So, entrepreneur Joe has such an idea. Then Joe is about one in one million. Now how will Joe do on Dodge's points?

Will Joe be able to get others 'excited' about his idea that only Joe understands? If they are as well qualified as Joe in Joe's specialty, maybe, but we already know that Joe is one in a million. So, initially, likely Joe will have to proceed alone.

Will Joe be able to get VCs excited about his idea? Not a chance! In information technology, one could count on one hand all the venture partners in the country able to evaluate an idea from a guy as rare as Joe.

So, continuing with Dodge's points, suppose Joe gets users, customers, revenue, earnings, all growing rapidly, that is, gets 'traction'. Then what? Sure, then Joe will be able to attract 'co-founders' and 'investors'. But then the question will be, "Just why would Joe want to do that?".

That is, with Dodge's criteria, for a good project like Joe's, by the time co-founders and investors are interested, Joe will no longer need them.

This is an old story that goes back to the Mother Goose story 'The Little Red Hen': She found a seed. From others she got only laughs. She attracted no co-founders or investors. She plowed, planted, cultivated, harvested, threshed, and ground the seed to make flour. She built a bakery. She mixed the flour to make bread dough. She let the dough rise and baked it into fresh, fragrant loaves of bread. Now she had customers, revenue, and earnings. Also now, and ONLY now, did she have co-founders and investors but needed neither of them.

But aren't co-founders and investors nearly always needed? Let's see: Let's start in the East in the US and take a survey of businesses, move to the west, and end up on the West Coast. Our survey will be in villages, towns, and cities. We will see some millions of businesses in auto repair, auto body repair, Web site design, grass mowing, roofing, kitchen remodeling, HVAC, dentistry, pizza carry-out, 'Diners, Drive-ins, and Dives', commercial and residential rental property, big truck-little truck businesses, etc. A huge fraction of these businesses are Sub-chapter S sole proprietorships and have no co-founders or investors. And at all the larger bodies of water, a huge fraction of the yachts are owned by such people.

The advantage of Joe? With a one in a million good idea in information technology, with the advantages of Moore's law, the Internet, and infrastructure software and the ability to execute that idea successfully, Joe just has some advantages over the guy with pizza shops, etc. Heck, a good commercial lawn mower now costs much more than a good server computer. It costs much more to start a carry-out pizza shop than a Web site serving 100 ads a second.

Joe is the guy to pay attention to. Dodge's advice is not for Joe and is for people not able to be Joe. To heck with Dodge.

NY_USA_Hacker··on Red Hot: The Computer Science Job Market
Do the jobs pay enough to buy two new cars and a nice house and pay for food, clothing, furnishing and running the house (insurance, utilities, maintenance), running the cars (insurance, maintenance, fuel), medical care, recreation, more insurance against risks, children and their education, and saving for retirement?

Typically a career will need to last over 40 years. How long with those jobs last?

When such a job ends, then what?

What these students will likely find is that the day they buy a house, assuming they can, a guy still in high school, with no college debt, started in plumbing or grass mowing, got his own business, hired a few workers, got customers in the neighborhood of the house, and 10 years later the computer guy got fired and was unemployable and the high school guy was able to buy the house.

In the US, being an employee sucks. Even being a computer science employee sucks. Instead, it is necessary to start, own, and run a successful business. The business might be just grass mowing or might be a hot information technology startup that gets a really good 'exit', but, whatever, it is necessary to start, own, and run a successful business.

As far as the "hot" job market, that's nonsense: The US is just awash in very well qualified people in software who are absolutely, positively, totally unemployable. The reason: They are over 35 years old.

Big secret: US hiring is still back in the days 80 years ago of a Ford factory where the supervisor was to know more and the subordinate was to know much less than the supervisor and there just to add muscle to the work of the supervisor. These college guys are 'desired' not because of what they do know but because of what they do not: The hiring managers assume that the college guys are naive, not very well qualified, not competition within the organization, and easily manipulated.

Here is a fundamental reason why: For a 'job', there needs to be someone else who creates it, that is, has the product or service, customers, revenue, building, office, desk and other equipment, other overhead, work to be done, and money to pay. It is rare for such a person to create a job good enough for the employee to have a good career for 40+ years and do well supporting a family and saving for retirement. Instead, the person with the chance of a good job is the one who created the job for the college guy. The guy with the best chance of a 'good job' is the guy who started, owns, and runs that business. For the college guy, that's the job he wants; that's the career direction he needs to keep in mind.

The real opportunity remains the one close to reality: Start, own, and run a successful business. For that opportunity, computing can help, but the field is not 'hot'. E.g., can also do well in plumbing, running several pizza shops, etc.

NY_USA_Hacker··on [2008] On The Flight to Old Text Editors
By far my most heavily used program is KEdit which is a PC version of the IBM VM/CMS editor XEDIT written by an IBM guy in Paris.

In both cases, there is a macro language based on Mike Cowlishaw's Rexx which is elegant.

Why such heavy use of KEdit? I am a heavy computer user, and, except for Web browsing, nearly all my computer input is via typing. So, the first issue has to be, what software to type into?

Well, I don't want to type into just a standard graphical user interface (GUI) 'multi-line text box' 'control' because the functionality is meager.

Yes, I'm a Windows user: So, in particular I don't want to type into Outlook or Visual Studio. Since my most important work is applied math, I use TeX for that and for all my high quality word whacking; so, I don't want to type into Word. I use SQL Server Management Studio for data base 'browsing' but won't type anything important into it.

So, here's much of what I type into KEdit:

(1) e-mail to send,

(2) input to TeX for mathematical and high quality word whacking,

(3) blog posts such as this one,

(4) notes on everything I want to remember in programming, shopping, equipment repair, medical care, financial records, phone numbers, addresses, abstracts of Web pages -- everything,

(5) software source code, including T-SQL for input to SQL Server,

(6) most of my important program test input,

(7) looking at program test output,

and more.

E.g., a little KEdit macro does my telephone dialing. Another little KEdit macro types mailing labels.

More? I'm developing a Web site; it's my first serious HTML programming. Since I'm a Windows user, I'm using ASP.NET. Well, I want some good logging for the Web site. ASP.NET has some default logging, but I didn't like it. Well, .NET also has some lower level tools for logging; these tools do handle the important issues of multiple threads of multiple programs on multiple machines all writing to the same log at the same time. Good -- that's the crucial functionality I needed.

So, now I have some good data logging.

Since the log file is 'circular', the first thing to do when reading one is to sort it on, say time, date, and machine. And with multiple log files, there can be some duplicate lines. In using the log files it can be important to 'select' lines based on the user's IP address, the name of the Web page, the routine in the Web page that wrote the line to the log, the message number of the message within the routine, etc. Sure, all this can be done with relational data base, but for log files of reasonable size it can also all be done more easily just within KEdit, especially if write a few KEdit macros.

So, I look at and do the first cut analysis of log files just within KEdit.

To be clear, on the title of the article 'On The Flight to Old Text Editors', I never left my "old text" editor. I've been using KEdit or XEDIT since 1985 and was using good text editors before then.

Alternatives? I see no good alternatives.

Why?

Well, the article lamented:

"That a new generation of programmers flocking to these old tools is concerning, if for no reason more selfish than the desire for peers in my dissatisfaction. Without a consensus that we can do better, there’s no incentive, no motivation, no market for improvements. If as modest a step towards a better editor as TextMate is abandoned, what hope is there for a true leap forward?"

Well, there's been no "true leap forward?" apparently mostly because of the overwhelming theme of the 'graphical user interface' (GUI) from the bean bag chairs and 'cognitive psychology' people at Xerox Palo Alto Research Center (PARC) and the use of GUI ideas by Apple, Microsoft, X-Windows, and now, with further development, smart phones.

In particular an important special case has been the Web and Web browsers.

The PARC goal was to make computer interaction easy for the pre-school children of the PARC employees for future Xerox office machines, and the approach was to have a user interface much like that on, say, a microwave oven.

So, with GUIs what we've got is for pre-school children using microwave ovens, and that's not promising for a "true leap forward".

So, the assumption in the computer industry has been that humans should interact with computers via a GUI, almost exclusively.

For a user mostly just receiving information, maybe that assumption is okay, but, for a user generating information, no.

The good text editors were designed by some of the best programmers ever for use by some of the best programmers ever. So, it's no surprise that good programmers prefer good text editors.

Bluntly, for generating information, it is still very important to type, and so far for people doing a lot of typing a good text editor is by far the best software to type into, in particular, much, Much, MUCH better than any of the check boxes, radio buttons, text boxes, and multi-line text boxes of the GUI standards.

In particular, for programming, I've been doing that for a long time. I've always found programming to be fast, fun, and easy giving the input just via a good text editor. I could count on one hand all the times I got any benefit from 'interactive debugging' tools.

So, I tried Visual Studio. I saw a window with many 'panels', and I could never find what all the panels were for. Some of the panels were said to be 'dockable', and the intended meaning was obscure and not in my copy of Webster's. The one little window was cramped, and in strong contrast when I program I typically have a dozen or so windows open at once, using either KEdit or my favorite Web browser, for entering source code, for documentation, for earlier versions of the program, for other programs I can use as samples, for various notes, for program output, for searching for information on Google, etc. The one cramped window from Visual Studio is, in comparison, a bad joke. For just the typing, Visual Studio gives me just some little panel within the one window, and KEdit is MUCH better. It's easy to write powerful macros in KEdit, and I do, and I saw no promising way to write macros in Visual Studio.

When I saw that with Visual Studio the work is to be a 'project' and even a project to report "Hello World" starts with 50 MB, I concluded that Visual Studio was from the far side of Xerox PARC in luna-land.

For my present development I'm mostly using just Visual Basic .NET via a script program from command line text windows. I'm thrilled: The complier is fast, easy to invoke, gives good error messages, so far appears to be essentially free of bugs, and generates surprisingly small EXE files. Terrific. What's in the Visual Studio 50 MB has to be glop, gorp, and goop I very much don't want to have to work with when things go wrong, and with that much glop, gorp, and goop some things are very likely to go wrong.

So, yes, I do a lot of typing and do my typing into the best text editor I know of, KEdit.

NY_USA_Hacker··on America Lacks Meaningful Innovation
The article is wildly wrong:

The Web? On the whole, it is wildly 'innovative'. Moreover, it is a huge aid to economic productivity and, thus, standard of living and quality of life.

Search engines? On the whole, they are wildly innovative and productive because they help solve a huge problem on the Internet -- finding stuff.

More in search engines? Yes, needed because the only search, discovery, recommendation, curation engines that work well are the ones based on keywords, and they work well for only about one-third of the content. How to get a search engine that also works well for the other two-thirds? That needs some 'innovation'!

A better search engine will likely have to make progress with how humans understand 'information' and 'meaning', and that will need some 'innovation'.

The author views 'innovation' mostly just in terms of what happened in the first half of the 20th century. Innovation is in different areas now. The author is also missing that to support the new approaches to innovation, Intel is now making 3D transistors with 22 nm feature sizes, and that in itself is history-making innovation.

Ads? So far the Internet is heavily supported by ads. Why? Because it's important to connect people with products and services. Why important? Because a key part of a higher standard of living is having people better allocate their limited resources, and key to this is getting people the information they need for such better allocations. One way to get the information is to have a person use a search engine to find the products and services. Another way is to have a vendor of products and services use ads to find the people. So far both ways are important.

The article wants to assume that ads are bad, and this is not good. There must be something very important about ads since they support nearly all of TV, radio, professional sports, just say, old media, and now the Internet. Pro basketball players? Basically they are in the ad business.

One reason entrepreneurs are rushing to Web x.0 companies is that's where the money is. And the money is not just in the US: The US companies that lead in this work rapidly become successful internationally. It's foolish to say that all this activity, internationally, is bad.

But not everything has to be international: The US did just fine, thank you, from about 1850 to 1950 growing mostly just internally and with relatively little role for foreign trade. So the claim in the article that there's something wrong with the US selling to itself is wrong: The better standard of living we want is from getting the work done, and we get it done by specialization with one person selling to another. All this does better when there is more efficiency, and the information via the Internet is a key here.

More information? Currently the US and more important world economies are in a Great Recession. The core reason? Bad information. Thus much of the solution will be better information. Generally information is just crucial for higher standards of living. The best thing that's happened for information so far is just the Internet.

Internet blogs are important, too: They can clear up misunderstandings caused by confused people!

But for some of the innovation the author seems to be dreaming about, that's coming: The foundation is Moore's law, the Internet, and infrastructure software. Broadly the path to more economic productivity is to automate, that is, have the machines do the work. Yes, we want the machines building houses and cars, growing crops, tending livestock, manufacturing boxes, bottles, and gadgets, making fiber, thread, cloth, and clothes, analyzing data needed for progress, yes, including in medicine, etc. It's coming.

NY_USA_Hacker··on Ask HN: End of IT Venture Capital?
Okay, to 'out do' the competition.

But that is a relatively new reason or emphasis if only because in the past 'venture capital' could invest in something on the back of a napkin and was needed for a lot in buying hardware and software, office furniture, recruiting, salaries, doing the development, building a marketing organization, starting the selling cycle, etc. before any revenue.

Now a motherboard for a 4 core processor with a 3 GHz clock is about $100; a server farm style 1 TB disk drive is about the same; Unix style software is not very expensive, and the Microsoft BizSpark program will let a startup use Microsoft software for free for a while, etc. The venture people want the development done, the site live, users, and, thus, likely revenue, and users and revenue growing rapidly. So, the situation now is much different than before.

So, before there were a lot of strong reasons to take equity funding, and now the reasons are much weaker.

NY_USA_Hacker··on Ask HN: End of IT Venture Capital?
Not from a good source, just from various claims that $1 to $5 CPM are common, some rough calculations starting with some summary data for Google, etc. I'd very much like a good source.

But if the $2 CPM is within a factor of several of reality, then the broad conclusion still holds: By the time a project is live and has users and, thus, qualifies for equity investment, if it is a good project with usage growing rapidly, then it should throw off enough cash to make equity investment unattractive due to how long it takes to get a check and the chances that after a year the revenue and pre-tax earnings will be larger than the check.

But, if cut the CPM by a factor of 10 to $0.20 and assume 5 ads per page sent, then the revenue per page sent is only cut in half and so are the monthly and annual revenue figures. Then the broad conclusion still holds: By the time the project is ready for equity funding, if it is a good project, that is, with usage growing quickly, then it should soon no longer need and maybe not want equity funding. And with some irony, here "soon" means in time about the same as that required to raise equity funding. So, the founders can sit there and ask,

"For the next six months, what do we do? Do we use the time to jerk the chains of a few hundred investors, pitch to a few dozen, and get a check or just keep working on our project, help the usage keep growing, and at the end of the six months have more in cash in pre-tax earnings than we would get from an equity investment check and still own 100% of our business and not worry about a Board. Also, now we just own the business outright, but the day after an equity investment we will own none of the business and, instead, have to work for about four years on a 'vesting schedule' to get back our fraction of the business, and we will never get it all back. So, what do we do for the next six months?"

So, broadly we have to ask, will there soon be some really successful Web sites that never took equity funding? So, for this part of investing, is venture capital dead?

NY_USA_Hacker··on 5 Things That Only Desktop Analytics Can Do For You (And Web Analytics Can’t)
.NET is not a "VM".
NY_USA_Hacker··on Kill Math
He doesn't understand math at all, not at K-12, college, graduate school, research, applications, etc. at all.

E.g., his claim about symbol manipulation is total nonsense.

His direction is a waste of time.

If he wants to improve materials for learning math, then fine, but he should first learn some math.

He should start with the books and papers of P. Halmos, one of the best writers of math ever.

Note: My Ph.D. dissertation research was on the math of stochastic optimal control.

NY_USA_Hacker··on My experience after pitching to an old school VC/angel
To play devil's advocate here, an investor SHOULD (although apparently few do) want to know about the 'potential' of the company after 500K users have signed up.

For this 'potential', not even the 500K means much. So, to understand this potential, what's an investor to do? One approach is to wait a few years, but then it's too late to invest. So, how can an investor understand in time to write a check to the entrepreneur? If I were writing checks, then I'd want a clear presentation in, say, a 'business plan' of just why the potential after the 500K users was so good.

Next, if I were an investor, I'd want this understanding of the potential after the 500K user BEFORE I wrote the seed stage check and before the 500K users had signed up. That is, I'd want before any checks were written a solid case made, on paper, that this was a good business direction.

I'd want to take that paper with that solid case, after the meeting with the entrepreneur, and study the case, think about it, etc. before deciding to write a check.

Actually, it really is possible to give good descriptions of significant projects just on paper and to do quite well evaluating the projects, again, just on paper. And, the evaluations can be so good that the subsequent execution can be quite reliable, even for leading edge projects. Examples? Well, there is a funder and user of technology. They are by a very wide margin the most important funder and user of high technology in Silicon Valley, the US, the world, and all of history in this solar system. Thankfully for US national security, this user is, yes, the US DoD. Some of their history includes WWII projects in radar, sonar, the proximity fuse, and the atom bomb. Then they did little things like the hydrogen bomb, ICBMs, SLBMs, GPS, stealth, the SR-71, the F-22, ARPAnet, aka, the Internet, little things like that. The evaluations were nearly all just on paper, and for approved projects the execution was good with very high reliability.

Evaluating projects on paper instead of just with 500K users can be important. Maybe you met the only VC left in the US ready, willing, and able to evaluate projects just on paper.

NY_USA_Hacker··on Peter Thiel Is Wrong About Higher Education (It’s Worse Than A Bubble)
I was going the other direction, and that is somewhat meaningful. That is, a well educated person, e.g., from Harvard, is skeptical.
NY_USA_Hacker··on Eureqa - detecting equations and hidden mathematical relationships in data
The claim that the fit gets to the "physics" of a pendulum is not well supported: The physics involves the law of gravity, Newton's second law, and aerodynamic drag as a function of velocity, and their work is very far from these three basic inputs from physics.

Indeed, a good intermediate step would be just to get the ordinary differential equation initial value problem for a pendulum, that really is from the physics, that their fit solves; they didn't do that.

Or if not a pendulum, then oscillations in basic A/C circuit theory with resistors, capacitors, and inductors; they didn't do that either.

Uh, data such as they started with ain't necessarily from a pendulum, and to assume that it is is a serious mistake. In particular there is no way that a computer program, or anything else, starting with just the data, can conclude a pendulum instead of an A/C circuit or something else. Sorry guys.

The role of 'complexity' in the expressions very much needs to be made clear. This is especially the case since otherwise can fit such data with just Lagrange interpolation (a polynomial) or splines.

The whole idea of the program encounters one of the biggest old problems in science: Starting with the observational data on the motions of the planets, Ptolemy had some complicated mechanical contraptions that fit the data fairly well (we will set aside some of the accusations that he, uh, 'smoothed' his data, that is, had an early use of Kelly's variable constant and Finkel's fudge factor).

The big problem in Ptolemy's work was his complicated contraptions didn't seem to be from any 'physical laws'. Of course, he didn't know that any suitable physical laws existed.

Big, huge steps later were from Copernicus, Kepler, etc. by which time it was clear that, looked at in the right coordinate system, the planets were moving in ellipses with the sun at one focus of each ellipse.

Then Newton assumed his second law and law of gravity, invented calculus, and used the three to derive the ellipses. That's why we think that Newton was one of the greatest of all scientists. The computer program is suggesting that it is hoping to automate such work; ROFL.

The claim that the program can find function f so that x = f(t, x) is not so amazing: Just set f(t,x) = x. Then perfectly x = x. Semi-amazing.

The example was for only one independent variable. More important cases are for several independent variables and several dependent variables. That is, for the real numbers R and positive integers m and n, find function f: R^m --> R^n so that for independent variables x in R^m and dependent variables y in R^n we get f(x) close to y. Pull this off with any generality, execution efficiency, and usefulness, and I'll start to be impressed.

To be more impressive than now, they might just look for linear f: R^m --> R and show why their work is better than old step-wise regression. There they need to address the conclusions now over 40 years old that step-wise regression didn't work very well, and doing all possible regressions took too long and gave a mess.

Generally they have the cart before the horse: The researcher is supposed to bring some understanding or at least some conjectures to the work and let the numerical work just fill in the details. Doing the numerical work first is bass ackwards.

The idea of a 'complexity measure' to select among different formulas is close to some old moon worship, black magic, or superstition about the magical properties of selected frogs and mushrooms.

Setting aside all the hype and nonsense, where could such work be used? Well, the answer is the same as has been the case for decades: It's a curve fitting program and, thus, could be used where curve fitting programs have been useful:

First, it can do the fitting arithmetic when the researcher brings to the computer some 'form' of the formula and is just looking for some constants and NOT the formula.

Second, if there is some data and want a simple expression that 'compresses' the data. That can be useful, say, in parts of stochastic dynamic programming; see, e.g., D. Bertsekas at MIT where he used neural nets for such 'compression'. That is, just storing the data might be 10^n bytes for n = 20, 30, 40, ..., but a few formulas might do just as well for big savings. Generally splines and multivariate splines have been among the leading approaches, but I didn't see splines mentioned in their video.

The scary, upsetting, objectionable parts are the claims that they are "detecting equations and hidden mathematical relationships in data" and, as in the video, getting at the "physics". No they are not. Here their hype reeks of the usual in 'artificial intelligence' and 'machine learning' that their software can 'learn' and be 'intelligent', and that's so far, and likely for a long time, nonsense.

Another objectionable part is their theme of setting aside all the assumptions, definitions, theorems, and proofs of statistical estimation of parameters in statistical models, hypothesis tests on the parameters, confidence intervals on the parameters and the predicted values from the fits, etc. How do they do that? They just leave it all out!

They have an option of 'smoothing' the data: That's a big subject. So, the implication is that researchers might just click on the button that says "Smooth" and report that in their work. Then to 'smooth' becomes some unlabeled bottle of snake oil.

The theme here is to omit all the rational and deductive framework of applied math and replace it with GUIs: This is a future of irrationality with no hope.

This work is a special case of the more general situation that 'computer science' is getting a C- in applied math and is out'a gas.

The work fills a much needed gap in the literature and would be illuminating if ignited.

NY_USA_Hacker··on How to: Pass a Silicon Valley Software Engineering Interview
"Statistical machine learning. Basically, the past decade of AI research has discovered that trying to encode a whole bunch of rules into a computer just doesn't work, because the real world doesn't function like that. Instead, what does work is to feed the computer a lot of data, and then notice patterns like "if this term appears, this article is most likely X". There're a bunch of techniques for this in the literature, they typically aren't taught in undergrad, and they are wonderfully useful."

Your first point is fully correct. The 'expert system rules' didn't work very well.

On your second point, "what does work", actually doesn't work very well either. You can use such things okay for first-cut, say, what J. Tukey called 'exploratory data analysis', then, to be followed by 'confirmatory data analysis'. The first part is okay, but the solid stuff is the second.

Then there's more: See what assumptions hold in the practical problem. Use these assumptions as the hypotheses of theorems with proofs. Use the conclusions of the theorems as results you now know about your real problem.

In real problems with a significant role for uncertainty, look for cases of probabilistic independence. For this, learn the high-end version of what a random variable is from, say, Cinlar, Breiman, Shreve, or Diaconis. Then for random variables X and Y, if knowledge of X doesn't help predicting the value of Y, then X and Y are independent, and can often come to this conclusion just intuitively.

Cases of independence can partition your problem into 'independent' pieces.

For more, look for some good examples of how to apply math in physics, chemistry, electrical engineering, and operations research. For statistics, sure, look at what agriculture did with the general linear model or, if you will, analysis of variance.

You will have little hope of making progress in using applied math just within CS or computing. For the progress, have to look to other fields that have made good progress with applied math. Eventually it will be crucial to have a good background in math, say, Halmos 'Finite Dimensional Vector Spaces', Rudin, 'Principals', measure theory and functional analysis, and then probability, stochastic processes, and statistics based on these.

NY_USA_Hacker··on How to: Pass a Silicon Valley Software Engineering Interview
What the grandmother did with ages, say, 0-99, would be radix sort if the sorter had 100 pockets. Then the sort was done in just one pass. I don't recall counting sort, but in this case it sounds like the same thing.

The obvious way to implement radix sort, just think punch cards, is stable with no extra effort.

NY_USA_Hacker··on How to: Pass a Silicon Valley Software Engineering Interview
First, let's talk about 'computer science' much as viewed via 'programming' and, indeed, as in the article.

Here's 'programming' in a nutshell: (A) Look at the problem and see what data is needed. Allocate appropriate 'chunks' of main memory for the data and give the chunks some names, usually mnemonic. (B) Manipulate the data with assignment statements from expressions, If-Then, and Do-While. (C) Otherwise use divide and conquer and call functions, subroutines, APIs, etc. So, that's the ABC's of programming.

For these ABC's all the common languages Algol, Cobol, Fortran, PL/I, C, C++, Rexx, Visual Basic .NET, C#, etc. are all very similar, more similar than English, French, Spanish, and Italian. 'Object oriented' programming? People did essentially this in Fortran and PL/I, likely also Algol, long before C++. IBM had object oriented programming in microcode in the early 1970s. Exceptional condition handling? Throw-Catch or Try-Catch are nowhere nearly as powerful as On Conditions in PL/I. Concurrency? We have mutex, semaphores, and actors and maybe more -- it's time for more: E.g., with lots of object instances and lots of threads, run into the same problems solved via transactional integrity in relational database with automatic deadlock detection and resolution, and basically, at least first cut, need the same solution.

Net, 'programming' as practiced and as in the article hasn't changed much in 40 years. In particular in recruiting no sense in getting all picky about some 40 years old material. The picky guy in the article didn't know enough to know how silly he was being. E.g., the corresponding guy at Google asked me what my favorite programming language was. I said PL/I. Of course he wanted C++. But, net, PL/I is a much better designed language than C++. My mention of PL/I ended the conversation.

Second, let's talk about what is important: Broadly, the future is important, and about that what is important is making it. So, how to do that? Here's by far the most powerful, broad 'paradigm': Have operations with powerful, known properties.

Now, considering how abstract computing is, to have some properties known is basically to have a theorem and a proof. Sorry 'bout that, but that remains the situation. Or, if we are going to build a large, complicated system where at each step all we have is "I think so, usually, maybe", then don't ask for much reliability for the resulting system.

Or, think like a mechanical engineer designing a Boeing 767: Start with some very well known engineering properties of the materials. From those properties use, say, finite element calculations, to determine the properties of various larger pieces of the airplane. Same for building bridges, buildings, cars, etc.

So, generally in building things, need to know the properties of what are working with. Also in computing.

Now, so far in computing, the more important 'pieces' were the now traditional topics in algorithms and data structures from Knuth's TACP, etc. Or we read about DeRemer's LALR parsing from Ullman. Or we read about k-D trees from wherever. Now ancient, but fine for 40 years ago and still good for the original purposes.

But for the future, we want more. Here's where computer science (CS) ran out of gas and got off track: CS wanted to move closer to the real problems. So, CS ran into 'information', 'knowledge', manipulating information, etc. and started borrowing ugrad texts in applied math, optimization, statistics, stochastic processes, graph theory, control theory, etc. These fields are based on theorems and proofs, but CS set those aside. Bummer since the theorems and proofs are the main way we know what we have, that is, the properties we need to depend on.

In particular, long the main criterion of artificial intelligence (AI) was just to program it and 'play with it and see if it seems intelligent'. So, the criterion was not that start with some assumptions that hold in the real problem, use the assumptions to justify some theorems, and use the theorems to construct the system so that we know what we have. Similarly there is 'machine learning' which also is working just empirically on the results and largely ignoring the assumptions, theorems, and proofs. Then there is the respect for heuristics, that is, guesses with no known properties except what we can see empirically. Basically CS and AI are becoming C- students in applied math, and that is NOT good.

My view is that the future of computing will have to borrow closely from the 'paradigm' of applied math, with a lot of careful attention to assumptions, theorems, and proofs. Then we will have 'pieces' with known properties we can assemble into the larger systems we need.

In a nutshell, CS and Silicon Valley know very well how to program at the level of ABC above but are very short on what to program. E.g., at Stanford, f'get about CS and listen to, say, P. Diaconis. At Berkeley, listen to L. Breiman. At Princeton, listen to E. Cinlar. At CMU, S. Shreve. At MIT, D. Bertsekas. Alas, none of these professors and none of their students would make it past the first-cut interview of the article!

Here's an example: Recently yet again we had a large server farm with lots of parallelism and redundancy go big time belly up. I mean, they had this backing up that, and when this got too full it got more from there, etc. Intuitively, it all looked really good. Intuitively. A proof? Nope. Was it good? Nope. It sucked. Terrible design that made the Titanic look good. Not nearly the first time for such a thing!

Let's just take the first, simplest part of this problem, near real-time detection that something is sick instead of well. So, we have two ways to be wrong, false alarms and missed detections. From Diaconis, we can learn about Neyman-Pearson (Neyman was long at Berkeley) where we see how to get the lowest rate of missed detections for whatever false alarm rate we are willing to tolerate. Now we are at the level of a junior course in mathematical statistics but already way, WAY past essentially all of the system monitoring community in both computer science and practical computing. NOT good.

Next, what we want, and we can't escape that, is essentially a statistical hypothesis test. So, we have data on each of several variables and with no reasonable chance of knowing the probability distribution. So, we need a multi-variate, distribution-free statistical test.

Now we are into reseach-level computer science. Just ain't going to get there with intuitive, heuristic, or empirical methods. Instead, notice where have some exchangability and a finite group of data transformations that are 'measure preserving' and derive a new class of hypothesis tests.

'Computer science'? Nice progress on an important problem in computing, but several chaired profs of computer science at top research universities and editors in chief of top computer science journals confessed that neither they nor anyone on their board of editors could review such math. Out'a gas.

For more on how to design a reliable server farm, need some more applied math, likely beyond essentially all of CS. Instead, for the right stuff as prerequisites, start with some good applied math people such as I mentioned. For CS, f'get about it. Right: For the future of CS, largely f'get about CS -- it's out'a gas.

The guy in the article doesn't have even as much as a weak little hollow hint of a tiny clue about this situation. The people he recruits are guaranteed to hold his company back to 40 years ago. He thinks he's a tough recruiter looking for the best of the best of the best, but he will only take people 40 years out of date. How 'bout that!

NY_USA_Hacker··on How to: Pass a Silicon Valley Software Engineering Interview
Now, the good answer, and essentially what the grandmother gave, is essentially just old radix sort like punch cards were sorted.

Radix sort? Take a deck of punch cards with a non-negative integer in base 10 right justified in some three columns. Run the cards through the sorter sorting on the least significant digit. The sorter puts each of the cards into the right one of 10 pockets, one for each of 0-9. Or just imagine that it builds 10 'linked lists'. Then move the cards back to the input, 0 pocket first, then the 1 pocket, etc. Then sort on the 10s digit column. Then repeat on the 100s digit column. Done. Just did the work in time proportional to 3*n for n cards.

If have a sorter with 1000 pockets, then can do the sorting in one pass. For the ages, such a sorter would sort all the input in just one pass.

Also cute, radix sort is 'stable' which means if card A starts before card B and if these two cards have the same value on the sort key then at the end card A will still be ahead of card B. Stable sorts are nice because they can permit sorting on several keys separately and, then, have the whole list of records in sort on the 'hierarchical key' made from 'concatenating' the separate keys.

Again, can do it all playing with pointers and linked lists.

It remains, for short keys, radix sort is faster than the (n)ln(n) sorts.

NY_USA_Hacker··on How to: Pass a Silicon Valley Software Engineering Interview
What you are calling "recruiting" I was assigning to HR.

If engineering wants an ad run, then HR can help as long as they don't touch the real content of the ad. When resumes come in, they can hold them up to the light, read them upside down, wave them with a magic wand, but under no circumstances should they make any significant decisions on the content. HR couldn't tell the next Turing award winner from a grade school video game player.

Who does college scouting for the NBA? College cheerleaders? Not exactly.

E.g., there's Gleason. That's A. Gleason. Since he never got a Ph.D., he shouldn't be a college prof, right? I mean, ask any college cheerleader! Instead, early in his career, maybe before he finished his Ph.D., he knocked off one of Hilbert's problems and was made a Fellow at Harvard and remained at Harvard for the rest of his career.

Just cannot hope to find Michelangelo to paint the ceiling if have house painters doing the first cuts on the applications.

← PreviousPage 2 of 6Next →