March 14, 2012
Thinking Machines

HAL 9000s malevolent eyeThinking machines come up a lot in fiction, and lots of people expect them to be realized some time in the future (any moment now, according to Ray Kurzweil and his ilk).

A corollary of the thinking machine is the simulated or uploaded human, produced by scanning a human being at whatever is the required level of detail and then computing a simulated world in which the scanned representation is evolved in accordance with the laws of physics. The content of the simulation could be interfaced with the outside world in some way, or an entirely self-contained virtual reality.

A materialist view of the mind seems to suggest that the simulation will have a consciousness (whatever that is), and if the scan is a good scan and the model is very precise it will be a lot like the consciousness of the original person. So far so uncontroversial. Imagine the second kind of simulation, one which is self-contained. The simulated person can’t be aware of the real passage of time; their conscious experience would be the same no matter how frequently the simulation was updated. Their experience occurs only during simulated time, which is not connected to real time.

In fact none of the underlying details of the computation are visible to the simulated person - if the simulator works correctly, all their experiments in VR should produce indistinguishable results from similar experiments in plain old R. For a lot of singularitarians and AI people, this still sounds quite reasonable.

If the thought experiment holds up so far, we can consider the next step: computing the states of the simulation in the wrong order. Remember that the subject has no perception of the passage of real time, only simulated time, which is only really marked by consistency within a single frame of the computation. Consequently, the order in which the states are computed can’t affect the conscious experience of the subject; nor can the computation of other arbitrary states in between.

At this point we have arrived at an odd conclusion: consider the following simple program

while (true) : increment counter

Given a large memory for holding the counter, this program will eventually cycle through all available memory states; if the memory were big enough to store a hypothetical simulation of the sort mentioned, all of the possible conscious experiences in that simulation must have occurred during the execution of the loop.

Note that we can also parallelize the implementation of this program very well, by using 2^n machines to “compute” all the memory states simultaneously. The fact that this is happening on different machines concurrently remains invisible to the simulatee!

In fact, we can add one additional layer of computation to produce an even odder conclusion; if we design our machine with nbit to have a memory controller which applies some permutation to 2^nbefore storing the state into the memory, each possible memory state will correspond to a different part of the computation of the conscious experience; however this effect will still be invisible to the simulated person (remember, they cannot see the simulator’s implementation details).

Allowing for this, it seems like a single large memory register initialized to zero also immediately contains all the possible conscious experiences of a simulation which can fit in that register; after all, the simulated persons involved cannot know which machine they are running on - the one machine is effectively simulating all of the different permuted machines at once!

I’m not sure what the logical consequence of this idea are; it seems like one of:

  1. The argument has a bug in it.
  2. Consciousness is not contained within a person, but somehow in an interaction between persons (i.e. the closed VR contains no consciousness, but a simulator talking to a real person does contain consciousness). This seems to lack neatness, and doesn’t fit the experience of being conscious.
  3. Simulations cannot be conscious, for some reason which is unclear.
  4. The argument is correct, and everything is conscious in all possible ways concurrently, everywhere and forever, but all these permuted worlds are closed off to us.

March 9, 2012
Legal Question

According to Mike Paynter, the law in Britain treats death threats less seriously if they have conditions attached; the case in point is one where this guy threatens to murder some individuals if they insult the prophet Mohammed. In this circumstance the maximum legal response is a £80 fine.

This leads me to ask what would would be the legal response if I were to make the following threat:

I am so enraged that I will kill you, if and only if every even integer except two is expressible as the sum of two primes.

The condition in this statement is definitely either true or false, but we don’t know how to show which (or whether it can be shown). Have I made an unconditional threat (large legal response), a conditional threat (£80 fine), or no threat at all (Scott Free)?

February 2, 2012
Conjecture about K-complexity

Conjecture: for any finite set of bit strings S and a chosen ordering of S, it is possible to define (or there exists) a machine M where sorting S by the Kolmogorov complexity of its members with respect to M constructs the chosen ordering.

Answers on a postcard.

December 12, 2011
"He has an honorary position as official Shoutsperson of the University of York’s Douglas Adams Society,[14] and in 2011 the University’s student body voted to name a new study space as the ‘Brian Blessed Centre for Quiet Study’.[15]"

Brian Blessed - Wikipedia, the free encyclopedia

(via gothick-matt)

November 12, 2011
(Untitled)

(Untitled)

November 10, 2011
The Social Graph is Neither

Lazyweb writes the article that explains what I think yet again.

November 7, 2011
(Untitled)

(Untitled)

November 2, 2011
Monochrome cupboard cat

Monochrome cupboard cat

October 31, 2011
The Mensch Hypothesis

You may have seen Louise Mensch being needled on HIGNFY recently; if you haven’t, take a look, it’s awful and funny at once.

Mensch suggests that there’s something wrong about the Occupy LSX protestors drinking coffee from the nearby Starbucks, something which renders their position invalid. Merton & Hislop both mock this view, and get some good laughs in the process; fair enough for a comedy show but this is the internet where such things are SRS BZNS. Let’s look at what Mrs Mensch says a bit more closely, as it’s a common refrain - “look at those guardianistas, smashing their system on their iPhones”, hilarity and dismissal ensue.

The argument here is one of hypocrisy, that lattes etc. are fruits of the economic system, and their consumption implies and depends upon supporting it. The protestors should be ignored because they are hypocrites (the cardinal modern sin), saying one thing and acting in contradiction. Let’s consider what is needed for this critique to hold up:

  1. The protestors believe they ought to be able to have a coffee
  2. The protestors take issue with the skewed wealth distribution and other current inequalities.
  3. (MH) Said wealth distribution is necessarily implied by the availability of a coffee. 
  4. Thus, the protestors’ actions (1) contradict their complaints (2) - they are not walking the walk.

From this we have the Mensch Hypothesis of capitalism: there can be no economy in which a coffee may be bought for a low price in which executive remuneration does not double every couple of years. No lattes may be had without the top 1% holding a radically disproportionate fraction of the wealth (and thus power). To shave even a single pound from the pension of Fred Goodwin or the salary of Bob Diamond would octuple the price of espresso, in all possible worlds.

November 22, 2010
[Flash 10 is required to watch video]

cat helping with work

10:17am  |   URL: http://tmblr.co/ZdOjWy1YB8VF
(View comments  
Filed under: cat helping