Monday, August 27, 2012

Gall's Law


I really like this law. Anyone know how true it is?

A complex system that works is invariably found to have evolved from a simple system that worked. The inverse proposition also appears to be true: A complex system designed from scratch never works and cannot be made to work. You have to start over, beginning with a working simple system.

It seems to say something like: 'You can't solve an optimization problem directly, you have to iterate'.

Although I can think of immediate counterexamples when it's stated like that, it brings to mind Galois' discovery that you can't directly solve a quintic equation by arithmetic plus extraction of roots. And so I wonder if that itself might be an example of a much more general truth along the lines of 'general solutions are always more complicated than the problems they solve, and so the direct solution to any sufficiently complicated problem is usually uncomputable except by approximation methods'.

So, can anyone think of any complex systems designed from scratch that worked?
Or complex problems that have easily computed solutions?

But actually Gall's Law says more than that. It says that if you've got a starting point of great complexity, your iteration process will work better if you give up and start again.

That again seems likely in the context of numerical solution of polynomials. Pick a random number as the start of your search and unless you're unreasonably lucky then the search will take longer than if you seeded the search with zero.

So, can anyone think of any complex systems designed from scratch which, although they didn't work to start with, quickly converged to working systems without going back through previous known working designs?

Thursday, August 23, 2012

Consequentialism FAQ

This is very good, if you like this sort of thing: http://www.raikoth.net/consequentialism.html

 I think I might be more sympathetic to consequentialism after reading it.

Wednesday, August 22, 2012

Brave 3D (Film)



This film in no way perpetuates cultural or gender stereotypes.

Fierce flame-haired daughters should be taken to see it, that they may learn humility and obedience.

Tuesday, August 14, 2012

Intelligence Explosion : We Already Have The Hardware



We have an example of an intelligence made out of matter, our own brain.

Whatever human-level intelligence is, it can be done on hardware of that complexity.

So it's worth asking what the complexity of that hardware is.

We think that whatever it is the brain actually does, the important level for understanding it is the neuron/synapse level.

Neurons look like transistors with thousands of legs and one output. Synapses are the connections between them.

The transistors that we know invariably have two legs. But we can chain them together to make a structure with a thousand legs and one output. We need a thousand transistors to do that.

So it could be that to make a working model of a brain, which implements the same algorithm, we'd need as many transistors as the human brain has synapses.

It's estimated that there are 100 000 000 000 000 000 synapses in a brain! (10^14)

Our largest commercial processors have something like 1 000 000 000 transistors (10^9).

So at first glance, it looks like we're five orders of magnitude short. That's a long way. It's the difference between a 1950s computer and a year 2000 computer. And we've got absolutely no way to predict whether the development of computers will continue at the astonishing rate that it's been going at over the last half of the twentieth century.

So most people who worry about the Singularity think that it's waiting on the hardware, and that the necessary hardware may never come.

But consider. The rate at which neurons can fire is something like 1000Hz. The speed of transistors in current chips is more like 3 GHz. That's six orders of magnitude.

It's always been our experience that making a computer ten times faster gives it more speed than making it ten times larger.

So I reckon that if we could work out what it is the brain is actually doing, then we could, on the sort of cheap hardware that lives in your desktop computer, make something that could do what the brain does, but at 10 times the speed.

And all my estimates above are highly conservative. Individual transistors can run much faster than GHz speeds. The speed limit for processor chips is to do with heat output and synchronization of the clock pulses across the entire chip. A neuron with 1000 inputs and one output probably can't do as much information processing as 1000 transistors.

And Moore's Law is currently continuing steadily. And whatever the brain does it is highly parallel, and nothing stops us linking processors together to get more total transistors, so if we knew how the brain worked and had a proper research budget to build a good one with, we should be able to make a brain that ran at least 1000 times faster than our own brains do.

But a lot of what our brains do is stuff like speech recognition, or vision. Computationally very hard, but not what we think of when we think of 'intelligence'. Whenever humans try to think about engineering problems, we run into having to do mathematics, or memorizing complex things. We are unbelievably slow at that sort of thing, which is actually computationally very simple.

At calculation, memory, symbolic manipulation, almost all the things that human intellectuals have traditionally found most difficult and most impressive, computer brains just blow us away by factors of billions.

These things are not natural for us. They take a great deal of training and practice. Consider the difference between a human child learning her own language and a human child learning the calculus.

Or your own ability to read emotions on strangers' faces versus your ability to solve simple probability questions.

This says to me that in order to do those things, we're in fact using mechanisms that were designed (by evolution) to do something else, and our schooling is a process of carefully repurposing and perverting these mechanisms to do new things, but to do them really badly and slowly.

A real computer brain could probably outspeed us by a factor of 1000 at the things that we're really good at, and that the great majority of humans can do very well, like catching cricket balls, or walking on two legs over rough ground, or imagining what is going on in other human brains, or finding our way home along a route after we've just walked a long way without a map, or recognising where would be a nice place to live, or throwing rocks, or throwing punches.

But at the things that we've traditionally thought of as our highest intellectual achievements, the things that very few humans can do well, which look more like accidental consequences of whatever our brains are actually for, things like mathematics and planning and logic and probability and scientific discovery, I would imagine that the speed factor would be more like several billion.

Luckily, we have absolutely no idea how the brain works, or what computations it performs.

If we knew how the brain worked, and what information processing it performed to work its miracles, and someone were fool enough to build a computer to do those computations, then they would have created a being which could recapitulate the whole of human scientific thought from Aristotle to Einstein in a few minutes.

That wouldn't really be something I'd like to play chess with. I've no idea how it would play. But I'm damned sure it would win.

A Slightly Harder Problem in Decision Theory


I'd be surprised if a cat could do this. But it think it uses the exact same tools as the easy one. 

The eccentric millionaire Oswald Mega walks into a bar and he says:

"This morning, I was showing my newborn about Dungeons and Dragons. We took a couple of six sided dice and rolled them, and wrote the results, which are just numbers from 2 to 12, on a piece of paper with 2D6 written at the top.

Then we took a twelve sided dice, and we wrote 1D12 at the top of a piece of paper, and then we rolled it lots and wrote down the results, numbers between 1 and 12, on the paper.

How she laughed at the difference in the patterns! Truly fatherhood is a joy.

Now, I've brought one of the pieces of paper with me, and if you can tell me which one it is, I'll give you £1000.

How much would you be willing to pay me to know the value of the first number on the sheet?"

And actually there might be some feline subsystem that solves a problem a bit like this.

Monday, August 13, 2012

A Deepness in the Sky (Vernor Vinge)



Vernor Vinge is a modern prophet. The inventor and populariser of the idea of a Singularity. Like Niven, a mathematician as well as a science fiction writer.

His most famous book is A Fire Upon the Deep. It's a clever and enjoyable book containing several excellent ideas, and it has a minor character in it called Pham Nuwen, who vaguely remembers the Slow Zone and the trading empire called the Qeng Ho.

Vinge's Universe is divided into regions, in order to avoid the modern science fiction writer's difficulty that even the very near future is likely to be incomprehensibly weird.

Gravity is opposed to Mind and Speed, so that the center of the Galaxy, the Unthinking Depths, is a place where no mind can exist at all.

The Slow Zone, which contains the Earth, can support human-level organic minds, and slower than lightspeed travel, but that's it.

Further from the core, you get the Beyond, where you can go as fast as you like and good computers are possible.

And after that, the Transcend, where intelligent minds become Gods and there is a Power known as The Old One, because it has been in continuous existence for ten full years.

I'm pretty sure that the reason for this scheme is that Vinge can't see how to set an interesting future story in the real laws of physics, where any technology much in advance of ours will just undergo Singularity and become a sphere of incomprehensible superintelligence expanding and destroying at the speed of light. He wants to play with these ideas but he has to keep a place where humans can exist in order to tell a story.

I was expecting the sequel, A Deepness in the Sky to be more of the same, to be honest.

But actually no, it's set entirely in the Slow Zone. It tells the story of Pham Nuwen and the Qeng Ho, and it tells this story as part of telling another excellent story about a single incident towards the end of Pham Nuwen's life, when he becomes involved in a first contact mission to an alien civilization.

The Spiders are the sympathetic characters here. They're very believable and actually quite interestingly alien, but also recognisably mid-20th century British. While the humans are the weird, monstrous lurking evil presence in space.

By the end of the book, you're completely rooting for the lovable aliens against the human space monsters.

The nice thing is that this is done whilst telling the entire story from the human point of view, and despite the fact that most of the humans are themselves sympathetic and likeable.

It's an unbelievable tour de force of shifting viewpoints and interwoven stories. Roles change around as the stories unfold and there are real and unexpected heroes and villians and victims.

This joins "Protector" and "The Sparrow" on my list of favourite books of all time. It's just much much better than its predecessor (and sequel) A Fire Upon the Deep, which is itself an entertaining, imaginative, thoughtful and important novel.

If I wrote science fiction, I'd like it to be like this. In the way that Sinatra sings the way I'd sing if I could. I'm now going to methodically hunt down and read everything Vinge has ever written.

Shangri-La Diet : Trouble in Paradise



Over the last couple of days, I've noticed a return of appetite. Nothing major, but I'm certainly remembering what it's like to feel peckish.

In particular, I've noticed that I'm actually looking forward to my couple of tablespoons of olive oil. Indeed I've been watching myself looking for excuses to have a third spoonful.

This is rather worrying. If Seth Robert's theory is true, then what that means is that I've managed to learn the flavour of the stuff and associate it with the extra calories that it's providing.

Shangri-La now predicts that it will act as a set-point raiser and my appetite should return in full force.

That means that Sainsbury's Mild Olive Oil is now just a normal food for me.

Oh well, off to the supermarket looking for oils that are unlikely to taste of anything at all.

Reading the labels when you know that what you're looking for is 'tasteless dregs ripped from the corpses of olives by unnatural means after all the good stuff is gone' is a most enlightening experience.

After a process of eliminating anything that claims to have any sort of taste, the candidates are:

Borges Extra Light Olive Oil
"True to the land and true to a long and deep-rooted family tradition"
"the result of an extraordinary blend of high quality refined oil and a slight touch of Extra Virgin Olive Oil"

I'm assuming that by 'slight touch' they mean 'homeopathic quantities'. It tastes pretty damned bland, but there's a slight nasty aftertaste.

Napolina Light and Mild Olive Oil
"Expertly blended for a softer and more delicate taste"
"composed of refined olive oils and virgin olive oils. comprising exclusively olive oils that have undergone refining and oils obtained directly from olives"

Ugh! This is actually bitter. If you put this on food it would make it inedible. What in the name of God do you have to do to olive oil before you're no longer allowed to claim that it has been obtained directly from olives?


Filippo Berio Mild and Light Olive Oil
"All the benefits of Olive Oil without a distinct olive flavour"
"composed of refined olive oils and virgin olive oils. comprising exclusively olive oils that have undergone refining and oils obtained directly from olives"

Notice that the second text is the same as the Napolina version. I wonder if they're the same company. If not, has one set of weasels actually stolen the weasel-words from the other set of weasels?

This is actually not bad stuff. It's very like the Sainsbury's stuff except even blander. Unfortunately I think I like it a bit too much. It's presumably the same taste as the stuff I've habituated to.

And finally for completeness:

Sainsbury's Vegetable Oil
"We're sure you'll love this product. If you don't simply return for a full refund"
"rapeseed oil"

I wholeheartedly approve of the lack of weaselling here. No one is even pretending that this is good for anything other than frying chips.

As far as I can tell, this tastes of absolutely nothing at all. I neither like nor dislike it. It is therefore declared the victor.

Sure enough, a hearty swig from the bottle is in no way a pleasant experience, and thirty minutes later I can no longer remember why anyone would ever want to eat anything ever.

I am changing my experimental protocol to read "Two Tablespoons of Sainsbury's Vegetable Oil" instead of "Two Tablespoons of Sainsbury's Mild Olive Oil"

I don't think I'm cheating here. Anyone disagree?

And does anyone want three bottles of nasty olive oil? Otherwise they're going on the common in the hope that something will enjoy them and that they won't kill the grass. No way is any of these horrors going anywhere near actual food.

The things I do for Science. I'm not even fat.

Friday, August 10, 2012

Intelligence Explosion


Recently I've been asking for help battling against an overwhelming complex of science-fictiony beliefs that I'm coming to hold. I hate having these beliefs, mainly because they feel uncannily like what I'd imagine a new religion must feel like to an early believer. It's very much not compatible with my self-image to fall for a millennial cult, and I trust that feeling.

So I've asked various friends to have a look at this belief-complex, in the hope that they can point out where it's a load of rubbish.

I'd actively like to be ridiculed here.

I once started feeling all spiritual after a particularly pleasant evening in King's College Chapel listening to Bach, and a few minutes of ridicule was enough to dispel the feeling and save me from the hideous mind-virus that is Christianity.

Unfortunately, the point in this set of ideas that people are consistently choosing to attack is the idea of an intelligence explosion. The same friend who saved my soul from Christ just looked at the idea of an intelligence explosion and said 'Nah, can't happen', and then stopped thinking.

The possibility of Intelligence Explosion is something that I've been worried about for many years, and it's not the bit that feels religious. It just feels like a threat that ought to be taken seriously. A bit like 'Earth might be Hit by an Asteroid', or 'Global Warming', or 'Nanotechnology Eeek!' or 'Nuclear War', or 'Flu Epidemic', or 'Worldwide Dictatorship', or 'Surveillance State', or 'Terrorists Who are Actually Good at it Rather than Mediaeval Halfwits with Towels on they Heads'.

i.e. It's a horrible existential risk, but there's bugger all I can do about it, and it's fun to think about, and it probably won't happen in my lifetime anyway.

I don't have children, I don't plan to, and although I care about you and your children, and hope for immortality myself, I don't care/believe enough to mean that I can't think straight about these problems and look them in the face.

And the idea of an intelligence explosion is just obvious in retrospect, isn't it? Like the idea of a chain reaction, or the infinity of the prime numbers. I'd never have thought of it myself, but once someone points it out, your head would have to be way broken not to take it seriously.

And it occurs to me that if I think it's that obvious, and I think that I am good at explaining ideas, then I should be able to make it obvious to other people. And so I'm going to try.

A Very Easy Problem in Decision Theory


Honestly, I reckon a cat could do this one, mutatis mutandis:

A barrell contains 25% diamonds, and 75% circles.

10% of the diamonds are blue, the rest are red.

80% of the circles are blue, the rest are red.

One of the objects is pulled out at random, and you're told the colour.

You can take a guess at the shape. If you're right, then you get $10, if you're wrong then you get a $1 for your trouble.

You do pretty well if you always guess circle. Can you do better?

In fact it could be argued that a cat is a mechanism for solving such problems, amongst others.

Tuesday, August 7, 2012

Is Murder Immoral? Is it Possible to Achieve Immortality by Not Dying?



It seems to me that if you're a materialist, then Cryonics is a pretty good bet.

When you die, you get your ass frozen in liquid nitrogen.

The argument that convinces most programmers that this might work is: 'this is not a secure delete algorithm'. If you take the hard drive out of a computer and put it in a fridge, then later on, you can read the data off it even if the mechanism's buggered.

If you want to make sure the data is not recovered, then you have to work a bit harder. If you put it on a bonfire for example, then that data is never coming back.

All the traditional methods of corpse disposal are bonfire-like. No way to reconstruct the information that was once there.

But cryonics will preserve the brain in the state it was a few hours after your heart stops, and at that point, we can imagine (with some difficulty) reading out the connections between the neurons. If that's all there is to being you (which is our best guess), then we can either (with difficulty) rebuild a brain with that wiring, or (easily) run a computer simulation of one.



And any good materialist has to reason: "I can have a conversation with this new person/computer program, and he will claim both to be conscious, and to be John, and be quite sincere in that, so I had better accept that he is John."

But of course, once we've done this, do we destroy the original body? Do we destroy the records?

What if, one day, long after your new self has headed off swanning around the Milky Way, someone finds a way to resurrect the actual body, rather than just the pattern? Which of these two people is you? What if they just make two copies of you? What if they slow down the clock on the computer simulation so much that the next scheduled tick is after the heat death of the universe? Have you just died again? Should you mind?

It looks like we're about to build a transporter. In our real universe, not our fictional one. Which of course is why I'm currently worrying about the morality of Star Trek. It's easier to think about, but the issues are the same.

It gets worse.

If you throw your hands up and decide that the future cop(y/ies) of you aren't you in any sense that matters to you (no one else will notice any problem), then consider the following:

When you go to sleep, you lose consciousness. Overnight, the configuration of your brain, the very atoms of your brain change much more than they would in an hundred years of cryonic storage.

If the cryonic you isn't you, then neither is the you that will be tomorrow.

Do you believe that you die every night, and someone else wakes up every morning?

If that's true, it's not so bad. We've been putting up with it for ages, and never suffered the slightest inconvenience by it.

If you give someone a drug that makes them sleep for days, have you killed them?

If you give them a drug that makes them sleep for years, have you killed them?

If you freeze them while they are asleep and wake them up thousands of years later, have you killed them?

If you freeze them while they are asleep, keep them revivable but never revive them, at what point did you kill them?

What the hell does it even mean, to be immortal?

Consciousness, we directly experience. No philosophy that does not account for it is acceptable to me.

Continuity of consciousness, not so much. Why should I care whether or not someone who isn't quite me, but who is quite like me, rises from the dead in 100 years time?

But then if I buy that and decide I don't care, why should I care if someone who isn't me, but who is quite a bit less like me that the revived cryonics patient would be, rises from my bed tomorrow morning?

So can anyone come up with an argument against cryonics (which costs around £30000 plus hassle) that doesn't also imply that if I thought someone was trying to kill me in my sleep I shouldn't mind?

Does Smoking Prevent Lung Cancer?

In 1964 the US Surgeon General released a famous report showing that smokers were much more likely to die of lung cancer.

This report came under attack from statisticians all over the world, since it had reported a correlation and inferred a causation.

The legendary Hans Eysenck, who coincidentally at the time was receiving a very large amount of research money from the tobacco companies, proposed that there might be some sort of hidden factor which caused both smoking and lung cancer.

This argument was greeted with the derision it deserved, and served mainly to discredit Eysenck in the eyes of the public, and cement the reputation of the tobacco companies as weaselly amoral murderers ( I speak as a happy customer of many years. )

But ridicule and cries of "come off it" shouldn't be enough to demolish an argument. I believe that they laughed at Christopher Columbus, when he said the world was round[1]



Can we make Eysenck's argument look good?

Well, suppose that there were a vitamin deficiency disease, associated with the currently undiscovered Vitamin N.

Primarily it affects the lungs. If you're deficient in this vitamin, your lungs don't work so well, and you have a tendency to develop lung cancers in later life.

As it happens, one of the hundreds of ingredients in tobacco smoke is chemically close to Vitamin N. When people with the deficiency smoke, they find that they enjoy it! A bit like if you haven't eaten fruit for a while, Vitamin C tablets taste extra-nice.

If this was the whole story, then we'd expect to see that non-smokers died of lung cancer much more often than smokers. Which we obviously don't see.

But what's a vitamin and what isn't is genetically determined. Almost all animals can make their own vitamin C. For some reason, the fruit-eating great apes, who have a lot of it in their diet anyway, have lost the ability. Humans are a type of great ape. So we don't have the ability either, but we've moved away from the diet of fresh fruit without regaining the ability.

Without a source of vitamin C, we die of scurvy. Luckily we can get it from fruit or fresh meat. If we start trying to live on preserved foods, without taking precautions, we die.

But there must have been a point in the history of the great apes where there were some apes who had the ability to make vitamin C, and some who didn't. It just didn't matter very much when they lived on fruit.

So suppose there's a gene for making vitamin N. Some people have it, and some don't.

Let's say that half of people have the gene.

They don't tend to take up smoking because it doesn't cure any lack they have, so they never 'get the taste'. Suppose 20% of them smoke. And they're unlikely to get cancer. Say 20% chance. But if they smoke, it does protect them a bit, so they only have a 5% chance.

But the other half of the people don't have the gene. Unaided by tobacco, they'll have a full 90% chance of spontaneously contracting lung cancer. If they're lucky enough to try smoking at school, then they'll find it very enjoyable, will likely continue to smoke all their lives (80% chance) and it will reduce their death rate to 80% (it's not a very good substitute for vitamin N, but it does some good).

What will the numbers look like in this world? ( Which I find all too plausible. Nature is always playing nasty tricks like this. )

Consider 200 people:

100 of them have the gene, 20 of those smoke, and 1 gets cancer. 80 don't smoke, and 16 get cancer.

100 of them don't have the gene, 80 of them smoke, of whom 64 get cancer. 20 don't, and 18 of them get it.

So if we don't know about vitamin N or the gene for it, and we do a cancer study, then

100 smokers, 65 cancers
100 non-smokers, 34 cancers

and the US Surgeon General issues his report, and up goes a huge public outcry against tobacco, and a few statisticians point out that correlation is not causation, and are ignored.


Exercises for the Reader:

This argument is utter bollocks from start to finish. But why?

Next time someone publishes a study saying that vitamin supplements shorten life expectancy, what will you believe?

The standard answer to this sort of question is a controlled randomized trial. Consider how the idea of an experiment where you ask people to take up smoking is likely to go down at the ethics committee meeting that decides whether you're allowed to do it.

No, you can't do it on dogs. Dogs don't get scurvy either. They can make their own vitamin C. That's why they can live on tinned meat and dry biscuits.


----------------------------------------------------------------

1/ Actually sailors have always known the world was round. They laughed at Christopher Columbus for badly miscalculating the size of the thing and thinking he could sail to India the long way round. The error should have killed him and his men, but they got lucky and found the Caribbean Islands instead. I believe that Colombus died thinking that he had found a nice short route to India. Jamaicans today immortalize this mistake by calling themselves West Indians.


Simpson's Paradox


Suppose that you went to the doctor, and he told you that you had cancer.

And he said that there were two possible drugs, placebifen and nogudox, which he could give you.

And luckily there had just been a big survey done, where they tracked the progress of lots of cancer sufferers, taking both drugs, and a higher fraction of people taking nogudox recovered.

Which of the two would you ask him to prescribe, do you think?

A while later, you're at the pharmacist, collecting your nogudox, and the chemist tells you that he has just seen a large survey published in the science supplement of the Wisbech Advertiser, where for males like yourself, placebifen proved to be the better drug.

And because this is the twenty first century, and the Advertiser has an online edition, you go googling, and actually it is worse than that.

In the WA survey, placebifen was indeed slightly better for males with your type of cancer, but it was also slightly better for females.

So now you are a bit confused. If you are a man, placebifen works better, and if you are a woman, placebifen works better, but if you are a person, then you should go for nogudox.

What on earth do you think is going on?

As fine a place as the internet is, it's probably time to talk to your doctor. He looks at both surveys, and looks a bit grave. It turns out that the Advertiser has published the same survey as he read in the Lancet, but they put the conclusions in the other order. (Obviously no-one ever troubles to read the bits in the middle. In fact many journals, to save on typesetting expenses, simply replace everything between abstract and conclusion with

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Donec a diam lectus. Sed sit amet ipsum mauris. Maecenas congue ligula ac quam viverra nec consectetur ante hendrerit. Donec et mollis dolor. Praesent et diam eget libero egestas mattis sit amet vitae augue. Nam tincidunt congue enim, ut porta lorem lacinia consectetur. Donec ut libero sed arcu vehicula ultricies a non tortor. Lorem ipsum dolor sit amet, consectetur adipiscing elit. Aenean ut gravida lorem. Ut turpis felis, pulvinar a semper sed, adipiscing id dolor. Pellentesque auctor nisi id magna consequat sagittis. Curabitur dapibus enim sit amet elit pharetra tincidunt feugiat nisl imperdiet. Ut convallis libero in urna ultrices accumsan. Donec sed odio eros. Donec viverra mi quis quam pulvinar at malesuada arcu rhoncus. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. In rutrum accumsan ultricies. Mauris vitae nisi at sem facilisis semper ac in est.

Here are the conclusions of the survey:

In the Lancet:

            survival rate
placebifen  26%
nogudox     28%


In the Advertiser:

for males only

            survival rate
placebifen  20%
nogudox     15%

for females only

            survival rate
placebifen  40%
nogudox     35%

Pretty clearly, placebifen has a 5% advantage as a cancer cure, whether you're male or female.

Just as clearly, nogudox has a 2% advantage as a cancer cure if you're either.

If you're anything like me, and you haven't seen this sort of thing before, then you're staring in disbelief at these numbers, thinking 'But maths just can't work that way'.

That goes to show just how bad our intuitions are on matters of statistics and probability. I've seen this several times before, and I just constructed this example, and there is still a little voice shouting 'It cannot be'.

Here are the  actual numbers from the survey:

males given placebifen:   lived 40  died 160
males given nogudox:      lived 15  died 85
females given placebifen: lived 40  died 60
females given nogudox:    lived 70  died 130

Check the percentages yourself.

Monday, August 6, 2012

Death is Bad

Nick Bostrom has just dropped this anvil on my head with extreme force with his short story:

http://www.nickbostrom.com/fable/dragon.html

Friday, August 3, 2012

Shangri-La Diet II: Details of Test and Predictions for Second Month


The variables I can measure are:

favourite belt notch (1-6, where 1 is fat and 6 is thin)
appetite loss (subjective, I'll say can be either perceptible or not.)

My current state:

belt: 3 (that's comfy in the morning and stays comfy all day. Any looser and my trousers head south, any tighter and I'm conscious of wearing it too tight. Although when sitting down for long periods the damned thing cuts into my waist, so I tend to remove it entirely.)

appetite: I usually seem to be eating a fairly light 'brunch' and then not wanting anything else all day.
This is unbelievable for me. When people suggest going out for dinner I either go for social reasons and eat small amounts, or sometimes I just can't face it at all. Twice during the last month I have refused to go for a curry with old friends I don't see that often. I'm fairly sure that I have never done that before in my whole life and it felt weird. I was keen on the idea of the conversation and the taste, and I think commensal eating is very important for social reasons, but I just couldn't face the idea of eating.
(Even more weirdly, I've gone off fags and coffee, my skin and hair seem softer than usual, and various joint aches that have been building up over a misspent middle age have cleared up. As if someone was oiling them! That's just got to be psychosomatic, hasn't it?)

My original state:

Before I tried this I was at belt 2 in the morning, moving to 1 in the evening, and my appetite was "normal for me" (i.e. I don't think about eating usually, and in fact sometimes forget to eat and am reminded by headache or weakness, but become ravenous in the presence of food and eat vast amounts.)

The actual predictions of my six theories for the end of the month of August I decree to be:

Helplessness (weight stays the same, some loss of appetite to compensate for oil):

     1  2  3   4 5 6
yes  1  25 50 25 1 1
no   1  25 50 25 1 1
(total 203)

Willpower (no appetite loss, weight gain)

    1 2   3 4 5 6  
yes 5 5   5 5 1 1
no  5 75 25 5 1 1
(total 134)

Shangri-La (appetite loss, further weight loss)

    1 2  3  4   5 6
yes 1 5 25  75 25 5
no  1 5  5  5   5 5
(total 134)


I don't know what the Ruth theories predict about appetite loss, but if it works like Seth Roberts says, then learning the new flavours should bring back my appetite, so:

Ruth I (appetite returns, rebound back to original weight, possible overshoot)

    1   2 3 4 5 6
yes 5   5 5 1 1 1
no  25 50 5 1 1 1
(total 166)

Ruth II (have lost weight for month 1, will now plateau for 2 and 3, hard rain to fall in 4)

    1  2  3 4 5 6
yes 5 25 50 5 1 1
no  5 25 50 5 1 1
(total 174)

Ruth III (continue to lose weight as before, expect plateau 3 and 4, rebound months 5 and 6)

     1 2  3  4  5 6
yes  1 5 25 50 25 5
no   1 5  5  5  5 5
(total 137)


And what I think I should currently believe is represented by: 


H40:W36:S9:RI5:RII5:RIII5

If anyone is still reading this and can spot any place where I seem to be trying to fool myself, I'd very much appreciate correcting. I have a horrible feeling that if I was trying to test aromatherapy as a weight-loss technique I'd be writing similar posts and believing similar things.

Shangri-La Diet II: Another Go with More Theories


I'm going to carry on with the crazed oil-eating for another month and see what happens.

Currently my belt likes to be on its third notch all day.

My plan is to drink two tablespoons of (to me, utterly tasteless) Sainsbury's Mild Olive Oil as soon as I get up, then not to allow myself to touch anything I can taste (including toothpaste) for the next hour (best accomplished by going back to bed for an extra hour). After that I'm trying to deliberately forget that I'm trying to lose weight and eat what I like, when I like.

I called the three models that I had Willpower, Helplessness, and Shangri-La, and they make the following predictions:

Willpower: Eating 300kcals worth of oil every day on top of eating normally should result in weight gain.

Helplessness: Slight appetite loss should compensate exactly for extra calories, no change.

Shangri-La: Severe appetite loss should result in further weight loss.

I also want to introduce a new theory, which I shall call "Ruth". Ruth on the basis of many experiences with diets thinks that any new diet will work for a while, but after a bit you'll put all the weight back on.

That strikes me not compatible with the willpower model. If it's just to do with how much you eat over how much you exercise, then once you've lost the weight there's no reason that it should go back on.

But it is compatible with helplessness. By force of conscious will you can make yourself eat less, but your weight will drop below your 'set point', and the pressure from your subconscious to eat will grow greater, until you break and start eating enough to get your weight back up to where it's supposed to be.

And it's also compatible with the theory behind Shangri-La. As far as I understand it, Seth Roberts also predicts weight loss from any strange new diet. Your set point drops in response to not tasting familiar tastes as often, but after a while you adjust to associate the new tastes with calories, and then your set point returns to normal. In fact it was noticing this effect in both rats and on himself that prompted him to start trying to break that link. As I remember he predicts that the time lag for learning the new flavours is around a week. I'm pretty sure that's not happening to me. If it was, I'd be looking forward to my daily olive oil dose by now. As it is, it still tastes entirely neutral. I neither like nor dislike it.

Ruth doesn't say how long the cycle should take. So I'm going to split her theory into three: Ruth I, which says that you should lose weight on a new diet for one month, and then the next month should return to normal. Ruth II gives you two months of weight loss and two months of regain, and Ruth III is the three month version.

I think Ruth knows what she's talking about: Apparently the only robust result in nutrition science is that any attempt at weight loss eventually fails and the weight goes back on.

So I'm going to mess with my prior (currently H55:W36:S9). Ruth's theories seem like versions of my Helplessness scheme, so I'm going to take 15 points from that, and split it three ways between RI, RII, and RIII.

So, my prior beliefs are now frigged to be (H40:W36:S9:RI5:RII5:RIII5).

I must say that this really doesn't feel like what I actually believe. After last month's dramatic success I have a strong hunch that Seth Roberts is right. But I'm trying to use Bayes' Theorem to keep me honest.

So my official rational beliefs are now that last month's weight loss was a fluke, and that either the traditional Willpower (eat less, exercise more to lose weight), or my own Paleo/Helplessness theory (as long as you avoid fast carbs it doesn't matter what you eat, your weight will stay at its set point, bad luck if your set point is too high or too low) are true.

But I'm now allowing a combined quarter chance to a combination of Seth Robert's Shangri La Diet theory and Ruth's rebound theories.

And because my actual emotional belief has shifted strongly towards 'Roberts is right', I'll need to watch myself very carefully looking for ways in which I'm screwing up my observations or my priors to fool myself.

Never trust a brain! I'll repeat "If it's true I want to believe it's true. If it's false I want to believe it's false." several times every morning in the hope that it will magically make the untrusted hardware I'm running on more sciency!

Wednesday, August 1, 2012

Shangri La Diet: It Works!

This post is getting a lot of hits, months after I wrote it, so I ought to qualify it. In the initial month of trying the Shangri-La diet, it seemed to work a treat. After that it got complicated, and there are several further posts on this blog about what went wrong and how I tried to get it working again.  

I certainly weigh less than before I tried it, but the amazing initial success lasted for exactly a month and then failed, in exactly the manner that habitual dieter friends have told me that any 'miracle diet' always does. 

After that, I had to hack it a bit to get it to start working again, and that required that I understood the mechanism by which it is supposed to work. I then experienced a couple of months of slow weight loss.

I am sure that this (modified version) works for me. Your mileage may vary, but I do recommend you try it. You'll know quickly whether it works for you, but if it does, watch out carefully for signs that you're getting used to the taste of the tasteless calorie source, and take appropriate action. The version that worked for me is:

Drink 300 calories worth of extra-light olive oil first thing every morning and immediately wash the taste away with water. Let nothing except plain water pass your lips for one hour after that. Apart from that, eat whatever you like, whenever, for whatever reason. It doesn't matter in the slightest if you miss the odd day.

Here's the original post:




Well, in a qualified way.

The favourite notch on my belt has moved one place smaller. I'm visibly thinner and feel it. And the appetite loss is undeniable. I've pretty much forgotten what it's like to be hungry.

I'm slightly annoyed to have found evidence for something so silly on my first attempt to run an experiment on myself, and I strongly suspect that I've done something wrong, but I can't figure out what.

There are various reasons to believe that I've fooled myself here, but I don't find any of them really convincing:

Primarily, I've smoked more this month than I was expecting: There've been many cricket matches to play, many birthdays to celebrate, an old friend died and there were two associated wakes, and the Town Bumps races led to entire week of heavy drinking.

So I've essentially been drunk for the entire month. When you're drunk your willpower goes, and this tends to lead to smoking. But I'm nowhere near back to my usual rate.

But this may well be the whole reason for the weight loss. If a mild increase in smoking works this well I'm surprised it's not recommended for health reasons. I mean, it's obviously deadly, but I think most people would swap ten years of old age for a lifetime of obesity. And I've never heard anyone recommending heavy drinking as a route to weight loss!

My second reason for doubting myself is that I found I wanted the Shangri-La diet to work. It's clearly ridiculous, in fact if you were wanting to come up with a parody Placebo-Diet backed up with crazy pseudoscience you'd be hard-pressed to do better than this. The only improvement I can think of is to use sugar-pills instead of olive oil. And that's apparently pretty much what the original version Seth Roberts came up with was like.

Originally I was interested in checking out a hoax that appears to have convinced many people despite a complete lack of rigorous evidence. But the minute I noticed the appetite loss I started to believe, and I found I wanted it to be true. Around half-way through the month I noticed that I could wear my belt comfortably on a tighter notch in the morning, and I found myself trying to wear it on that notch throughout the day even though it became uncomfortable.

However, I've noticed this effect, and I've tried to counter it, using the defence that Eliezer Yudkowsky has called the Litany of Tarski: "If this works then I want to think it works. If this does not work then I want to think it does not work. Let me not become attached to beliefs I may not want."

And I'm pretty sure now, after 30 days, that it's moved slightly more than a whole notch. My belt is comfortable on its third notch in the morning and stays there all day now. At the beginning of the month I wrote that it was notch two in the morning and needing loosening to notch one as the day wore on.

So I'm calling this as a qualified success for Seth Roberts and Shangri-La. I'll update as planned so:

The odds of this result for the Willpower theory I guessed at 5/126. for Shangri La 50/137. For Helplessness 5/126

And I took my prior to be H60:W39:S1

So I update to H 60*5/126: W 39*5/126 : S 50/137

Multiplying through and rounding off: H55:W36:S9

And suddenly I'm shocked! This doesn't reflect my current beliefs at all.

I've tried this new thing. It clearly works. Emotionally I'm convinced that it works, and I'm going to keep trying it for the next month and I now confidently expect that it will continue to work.

Bayes is telling me that I don't have enough evidence to update my beliefs like this.

Given how unlikely I thought this theory was a priori, the fact that my test has landed on its central prediction should most likely indicate that some other theory is true, but that my weight loss happened by accident.

I can completely buy that now I've thought about it. The helplessness and willpower theories have had some of their probability mass stolen by the mad Shangri-La theory, but a one-person test (even if I had made accurate priors and controlled outside factors, which I didn't even attempt), is nowhere near enough to give Shangri-La an edge.

In fact the sheer dodginess of my one-person trial, and the placebo effect and various other expectation effects mean that I shouldn't update even this much, but if I had found no effect and refused to do the full update then I'd accuse myself of being biased, so sod that.

So I'm going to try to believe that there's a ten per cent probability that the Shangri-La Diet works. I don't think my brain's really set up for that!

Followers