Re: How to make money fast (in science fiction)

In rec.arts.sf.science CharlesRCaplan@[EMAIL PROTECTED]
 wrote:
> Sure, on the other hand he could go for the big money and write games.
> Or just write the graphics engines that run games. Assuming he's super
> intelligent, there are lots of ways that graphics/physics engines can
> be streamlined to operate faster and be easier to interface with. If
> he could write directly in machine code, with perfect algorithms, and
> flawless execution he could produce noticeable gains in performance.

He *may* be able to write better engines from an internal standpoint, but 
I'd guess that he won't be able to produce any generalized engine usable 
by others. The reason being that he'll be too smart to produce a useful 
API.

You see this kind of problem all the time now. A programmer produces a 
library, and the first attempt is useless to anyone but him. The reason is

because the API he exposes tends to assume that the user is as fluent in 
the problem as the original programmer, simply because the original 
programmer is too close to the problem. Having an enormous excess of 
intelligence doesn't strike me as something that will solve this, but 
rather the opposite; he'll have an even harder time than usual trying to 
get down to the level of the engine users.

> Ever wonder why computers are something on the order of thousands of
> times more cycles per second and have thousands of times more RAM but
> still don't work much faster than they did in 1995?

I must object to this. My computer is vastly faster in basically every 
conceivable way than the one I had in 1995. It "boots" nearly instantly, 
it is almost instantaneously responsive to everything I have it do, and 
when it isn't it's because I'm having it process more data than the 1995 
computer could even conceive of dealing with.

Not to mention that for a factor of "thousands" you need to go back 
significantly before 1995. A reasonable computer in 1995 might have had a 
100MHz processor and 8MB of RAM. An equivalently positioned computer today

might have a 2GHz processor and 2GB of RAM.

> Software bloat.
> Applications become more complicated but programmers don't. The newer
> programming languages and development environments become more
> complicated allowing the programmer to produce more capable software
> but at the cost of massively bloated code underneath.

I must object to this as well. Many programming environments currently in 
use are effectively fancy macro assemblers (the ones built around C-based 
languages). The ones that aren't are often put through fancy compilers 
which do a very good job of optimizing around the extra stuff. For the 
most part, software seems slower because it actually *does more*. A word 
processor from 20 years ago would be rendering bitmapped fonts with 
pixel-granular spacing and showing you maybe a paragraph or two at a time.

A modern word processor will show you a couple of pages at a time, 
rendering vector fonts with subpixel antialiasing and layout, all the 
while doing dynamic spelling and grammar checking, and it's *still* vastly

faster. On a code-per-capability basis, I see no evidence that modern 
software is all that much worse than it ever was. It certainly is somewhat

worse, simply because people were constrained to spending lots of time 
squeezing their code into the hardware available rather than spending it 
on making it more useful, but this is probably a factor of two or three at

best.

> If this super
> intelligent guy could write directly in machine code with perfect
> algorithms and able to manage all the complexity that something like
> this would entail, then he could write his own paycheck. Assuming that
> he kept it a secret how he does it.

Very much unlikely. Nobody in their right mind would ever hire a guy who 
works like this. Someday this person will quit or be fired or be hit by a 
bus, and when that happens you will need his replacement(s) to be able to 
maintain his code. If he wrote it all directly in machine code while 
managing the complexity entirely within his head, then the moment he steps

into Main St. without looking and ends up in the city morgue, you may as 
well delete everything he wrote and start all over.

Not to mention that this technique still isn't going to help that much. 
Nearly all software follows an 80/20 rule or even a 90/10 rule when it 
comes to performance. That is, 10% of the code takes up 90% of the 
computer's time.

Putting much effort into the 90% of the code that only takes up 10% of the

time is pretty much wasted; even if you get its execution time down to 
*zero* you'll still only improve your overall performance by 10%.

Putting effort into the critical 10% is where you make your gains, but of 
course everyone knows this already and puts a lot of work into it now. 
Certainly the product isn't perfect, but I would wager that for a 
performance critical application with a lot of development resources 
available, they're getting to within at least a factor of 2-3 of the 
ideal.

The story of the lone genius programmer who can arrange things the way 
nobody else can and who produces amazing but incomprehensible code is a 
nice one, but it's a myth. It may have been true at one time, but anybody 
like that today is going to have a difficult time holding a job at all, 
much less "writing his own paycheck".

> Operating systems is a dead end though. Microsoft products are some of
> the worst around, there are other better ones available but MS is
> still on top due to entrenched market share.

I'm not convinced this is the case either. The two real competitors to 
Windows are Linux and Mac OS X. Linux is still a pretty terrible OS for 
anyone who isn't a professional computer person. Not surprisingly, in 
areas where the only users are professional computer people (such as 
servers, databases, embedded devices, etc.), Linux has beat Windows fairly

convincingly. Mac OS X is a really nice OS for the average Joe but is 
deliberately tied to niche hardware.

> Though if you could write a better operating system that also runs
> Windows software natively and seamlessly for the user, you'd sell a
> billion units. And you'd also get sued by Microsoft.

You may get sued but you would win. At least in the US, reverse 
engineering for the purposes of interoperability is explicitly described 
as legal even though reverse engineering is generally legal anyway. Such a

lawsuit wouldn't have a leg to stand on as long as the production of the 
better OS was done properly (i.e. no code or other things stolen from 
Microsoft) and the case should be so clear-cut that you should have no 
trouble getting money for enough lawyers to fend off the opposition.

-- 
Michael Ash
Rogue Amoeba Software