Friday, August 23, 2013

Self-Driving Cars Save Lives

The "self-driving" car is on the way and coming faster every day. It can't happen soon enough.  Car accidents kill millions, and maim millions more. Computer-driven automobiles won't drive better just because they drive themselves.

Self-driving cars will also be interconnected and constantly communicating with each other.  They'll "know" their own goals and intentions as well as those of the vehicles around them. Being able to rely on each other to the degree of effectively perfect shared understanding will enable them to plot safe trajectories and prevent nearly all accidents.

Even if someone should dart out in front of traffic, such as a child running for a ball, the cars will be able to react in ways and at speeds that are impossible for humans.

Cars as well as sensor nets in the road itself or built into local infrastructure would be able to monitor conditions and report them to the shared "road database", allowing cars to reduce speed near playgrounds or route around them altogether during games, for example.

Ubiquitous computing, sensors, and networking will make most accidents a thing of the past. They will also prevent such soon-to-be-anachronistic things as traffic jams and car chases.  You won't get moving violation tickets any more either. Driving will be like watching a really comfortable but slightly boring movie rather than the often-nightmarish gamble you take with your life every time you take to the road.

It can't happen fast enough.

It will also be a totalistic surveillance system of such intimacy and perfection as to constitute a synthetic Akashic records. In the process of driving these machines will know everything about everybody around them, and they can be ganged together to track everything they see, and they remember.

This is the price of the universal machine: universal transparency.  We are entering the time when we can make ourselves live to whatever rules we will, so it behooves us to do our best to make those rules indeed liveable (autocorrect suggested "lovable"...  Not bad machine, not bad.)

You are going to be surveilled. That cat is not going back in the bag.

You must make your peace with that, as I did.

But now we will live without hypocrisy (if we can!) and we will no longer mash ourselves up in our car crashes, and kill people just to get from point A to point B, and waste hours jammed in traffic.

We will have to learn to forgive ourselves for a lot of the shit we do, and learn to stop doing the worst of it, and pay for the bits that must be paid (the Devil will have his due, and some of us are very right to be terrified of the panopticon: I almost feel sorry for the ones who are only now realizing that the National Security Agency already has their dirty laundry.)

We will have to learn to be good. There is no other way to get good government. No matter the system it is made out of us, we rule ourselves whether we own it or not.  Yes your privacy is a fiction, but you might notice that suddenly most kinds of crime can no longer be committed!

We can actually detect most kinds of crime already from the existent systems we know about. I am convinced there are much more powerful and extensive systems still hidden, but the results of these systems can be selectively fed to outer levels for reduction of undesirable aspects of civil society. We should not be concerned that this will happen (as there is no way to stop it in any event) but rather that the definition of "undesirable" may be warped by residual power bases.

In the great struggle to govern ourselves the executive power has become perfected by the fulfilment of the Analytical Engine's great promise, and now we can rule ourselves completely for the first time since the fall.

Tuesday, August 20, 2013

Python as Basin


TL;DR: I am saddened but resigned to not hacking on tree because it's written in C.
I've often said that the only thing wrong with Python is that it ruins you for other languages.

What I mean is that nowadays, having written software mostly in Python for the last few years, when I go to read the source code for a simple tool like the 'tree' program (which reads a directory structure and outputs a simple text-mode graphical "tree" representing it) which is written in C I am struck by the amount of work I would have to do to be able to consider myself a "C programmer". (I used to be unembarrassed to claim "I can write C code." No more.)

Using Python has made me become monolingual.

I'm not suggesting that Python is "the best language" or anything (after all, we all know Forth is the best language...)  I'm just comfortable leaning on it heavily and neglecting other languages.

This is an old folk's liberty.  You young coders better learn as many languages as you can, it's not as hard as it seems and it's the only real way to escape the nest of abstractions that ensnares the unwary specialist.

(Put it this way: Yes there are more Java programmers than Python programmers but don't let that prevent your shop from deciding to use Python because a programmer who knows Java but can't pick up Python should not be hired! That's a baaad programmer! I would never dream of hiring someone to write software who knew Java (or C++) but balked at learning Python.  That poor bastard can go polish his turds on someone else's dime, Real Programmers only here please!)

So if language-agnosticism is so important to identify good programmers why am I so comfortable "degenerating" into monolingualism?

Well, for one thing, the characterization is false: I am still reading about and learning new languages. I just tend not to use them (other than for toy programs to play with them.)

At this stage in my trajectory as a programmer (which is much deeper and more important to me than my career as a programmer) most languages won't have much to teach me.  They also wouldn't enable me to achieve things at work that I otherwise couldn't achieve [with Python].

They are still fascinating. (I read "The Go Programming Language Phrasebook" cover-to-cover, drooling, just for kicks.)  But the romance has worn thin.

I want to get things done and I can do that with Python faster and with fewer errors (and fewer trips to the documentation) than any other available language (plus "ecosystem".)

There are sweet spots though when it would make sense to us another language.  If I had to write high-performance low-level software (instead of the endless succession of poorly-thought-out Django apps that have been buttering my bread for the last few years) I would lunge for Go.  It is a wonderful language and seems to me to be very well-designed.

I am less sure of the niche (for me) of Rust or Dart or (shudder!) Coffee script.

What I think we are experiencing in the field right now is a kind of shift from programming by language to programming by concepts.

The process of translating a high-level language into efficiently-runnable form is already so intricate and esoteric that it pretty much must be left to the crunchiest and most capable of us.  To put another way, it is very unlikely that a newbie is going to make significant contributions to a byte-code or assembly optimization process, but all-too-likely that newbies will inflict new languages on us (e.g. Ruby and PHP, both created by people who really had no business trying to create new programming languages.)

To me, this indicates that the "innovation-space", if you will, for PLs has to do with expressing ideas clearly, rather than some spectacular new syntax or nifty whiz-bang feature.  It is no fun watching languages slowly recapitulate LISP.

Let your language describe what you are thinking and let the accumulated wisdom and sweat and tears of the brilliant people working on translation and new hardware design turn it into efficient, runnable form.

In this regard Python is very satisfying.  It is possible to write very elegant and expressive code in Python. Readability is always the first thing on the list when I'm enumerating Python's strong points.

The PyPy project is beginning to provide a robust and powerful way to use Pythonic code in ways that are very flexible without sacrificing rigor.  (This is not to say that similar efforts aren't proceeding for other languages and "ecosystems"!)

I'm looking forward to the day that PyPy can convert my Python code into something like native code. (And of course there are tools like Cython.)

I only have so many hours in a day to devote to my craft, I can only burn so many calories to produce thought-stuff.  It made good sense to learn and use several languages when I was "coming up".  But now that I've "arrived" I find it makes more sense for me to improve by means of deep mastery of a particular set of powerful tools, rather than continuing to grok yet more variations on what I already know (syntax of languages.)

Learning Category Theory is more likely to pay off, personally and professionally, than learning Rust or Julia.