From: Dan Fabulich (daniel.fabulich@yale.edu)
Date: Wed Mar 29 2000 - 18:59:03 MST
'What is your name?' 'Zero Powers.' 'Do you deny having written the
following?':
> >*How do you know which screen to look at?* There are hundreds of
> >thousands, if not millions of cameras you could be monitoring, any of
> >which may have a picture of somebody slitting a child's throat. The only
> >way to tell which one is to LOOK at all of them, requiring a gargantuan
> >amount of processing power/time. That's where the computation comes in,
> >and a hell of a lot of it.
>
> *Which screen* you look at depends on *what* you're looking for. If I want
> to find out what *you* did today, I command "play Fabulich video 03-29-2000,
> 6:00 to 18:00 PST" and your boring day plays out on my monitor for 9 hours,
> or until I get bored and press "stop."
You're DEEPLY missing the point here. Sure, if you know when and where to
look, you can see what was going on there. How did you know that you
needed to look at *my* video to find the murder? How did you know when to
look? How does the computer know where I was? All of this requires a
thinker/computer to stare at the screen and do some processing. Without
it, you can't even make the above simple request for "Fabulich." On the
other hand, you COULD request "Yale University, undergraduate room 472B,
03-29-2000, 6 to 18 EST", but it's not at all obvious how you figured out
to look THERE knowing only that you wanted to find a murder.
> >So no one, not even a despot, could prevent you from
> >observing his every move. One would *still* need to analyze the data, to
> >figure out which screen to look at.
>
> Again, which "screen" you look *at* depends on what you are looking *for*.
> Think of the web as a huge database (which in effect it is). Right now
> there are millions and millions of pages on the web. But you don't need to
> do any big time "data analysis" to use the web. Why? You go to your
> handy-dandy search engine, type in what you are looking for and viola! a
> nice neat little list of choices pops up for you to choose from. I can't
> believe this is such a hard concept to grasp.
A very large amount of preprocessing has gone into the Internet:
everything you can search for there has been painstakingly typed in by a
human in a form that is easily searchable. It's already stored as
information which the computer knows to be text.
You may not be aware of this, but it's very difficult to get a computer to
correctly recognize that a picture of a given page of text is text at all,
say nothing of figuring out which text. You can search a copy of the US
Constitution that I type in and put online. You can't search a photograph
of a printout of that same document without a lot of processing and a lot
of hard work on the part of programmers. It's taken us years to get to
the point where we are today with optical character recognition; the
process has been extremely slow-going. Same goes for speech recognition,
the process of identifying a sound to be speech, and determining which
words are being said.
But identifying text is a piece of cake compared to identifying people.
Trying to explain what a person looks like to a computer is, as Eliezer
might say, just one step above trying to explain it to a rock. Today, we
don't have anything which can do this anywhere near as well as a human can
(to the best of my knowledge; this project is hard, but not impossible).
If you'd done very much programming you'd see why this is so hard. All
the computer knows is a grid of numbers (a grid which, as we happen to
know, corresponds to a grid of colors). Based on that alone, the computer
has to identify what a person looks like. Notice also that the computer
can only use simple arithmetic and set theory to process those numbers. So
I want you to imagine a function which uses only arithmetic/logical
operators, (+, *, =, OR, NOT) and membership relations, (X is a member of
set Y), takes as its input a grid of numbers, and yields as its output a 1
if the grid of numbers corresponds to a picture of a person, and a 0
otherwise. Can you see how hard this is? It's not impossible, obviously,
but it's Very difficult. That's why we still haven't done it thus far.
So the concept "is the letter H" and the concept "is a picture of a
person" are fairly complicated concepts which are very difficult to
explain to a computer. Now try explaining to that computer what
'wrongdoing' or 'harm to me' looks like in terms of the membership
relation and the successor function. Or what any action looks like.
Computers today can't reliably tell the difference between a video of a
man shooting a gun at a bunch of helpless innocents and a porn clip.
This is not because they're not powerful enough; if they were ten million
times faster, it wouldn't help. We have no good way of explaining to the
computer exactly which data in corresponds to which data out.
This IS a VERY hard problem. It's a problem which we cannot solve today.
"OK," you might say. "The AI of tomorrow will solve it. It'll be very
powerful. It'll use all the latest buzzwords." Fine. But that's a lot
of processing power you're talking about. Whoever controls all that
processing power gets to keep the results.
> Go back to the web analogy. The government doesn't control the search
> engines now. There is no reason that the "government" would have to control
> the transparency servers I talk about. Sure there could be government
> servers, but those would not be the *only* ones. Given the importance of
> the integrity of the data in such a society, redundancy would be the word of
> the day. Probably the best solution would call for an oversight body
> containing both government officials and civilians, like police review
> boards. Reliable checks and balances systems are not all that difficult to
> implement. After all here in the US we've been doing them for over 200
> years now.
Could happen. Nobody (except maybe Lorrey) will argue that if the system
you describe is implemented, it's guaranteed to result in a totalitarian
state. But every technology which benefits the powerful more than it
benefits the weak makes this MORE likely. Again, my argument is that
since this system makes despotism easier, more likely, etc. we should, to
whatever extent possible, avoid it.
The gov't doesn't control the web servers today, but it has no reason to
do so. It WOULD have a reason to try to seize control of servers of the
kind you're describing. And it would know if you ever tried to resist
them.
Checks and balances can work, but if the price was right, a despot would
throw them out the window.
> But, once again, regardless of who was in control of the servers, as long as
> the system was sufficiently transparent it would not matter. All that is
> needed is that the transparency of the process can be assured (and
> verified).
Look, allow me to provide a theoretical account from first principles. In
order to stop wrongdoing, you need all of the following:
1) You need to have the information that wrongdoing is going on. 2) You
need to recognize that wrongdoing is going on, and determine relevant
information about it (where it's happening, who's doing it, what you'd
need to do to stop it, given the power to do so, etc.). 3) You must have
the physical force to stop the wrongdoing, or you must be believed to
have the force to stop the wrongdoing.
In the Very Extreme Case, consider the situation where I've got guns and
shields and 1 and 2, whereas you only have 1. You MUST agree that in
*this* case, despite the fact that the scenario is perfectly transparent
("look ma! it's transparent BOTH WAYS") there isn't jack shit you can do
about it.
Now, you know that YOU'RE opposed to despotism, but do you know that
others are also opposed? Probably not. Why? Because the gov't in this
scenario has 1 and 2; it knows when you're trying to find subversives; it
knows what that looks like and can identify them instantly. You know that
it knows this. You don't know whether your neighbor is willing to revolt
or whether he's too scared to do so. You also know that the last time
somebody asked somebody else whether they were willing to revolt, the
government had them shot. Nobody found out until it was too late, because
they had 1, but not 2, which would have allowed them to find out in time.
The point here is that having 2 when everyone else doesn't is a huge
advantage. Having 2 and 3 when everyone else doesn't gives you the power
to be a despot.
> Keep the Internet analogy in mind. *Tons* of data, almost no data
> crunching. What you want is what you get, when you want it. Why
> would you think our ability to sort, organize, store and retrieve data
> will get *worse* in the future than it is now?
It won't. Video data is much harder to process than text. That's what
will make it harder.
-Dan
-unless you love someone-
-nothing else makes any sense-
e.e. cummings
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:27:44 MST