The fine tuning argument

[Pixel42]

The best response to the fine tuning argument I've ever seen is that it's the equivalent of me marvelling at the fact that not one of my ancestors, right back to the first cell, died childless. I mean, what were the odds against that?

 

[yy2bggggs]

Is this based on not getting a run of 4 or more?

No, it's based on the fact that you maxed out at 3, and that you only had 3 runs of 3 or more.

That is not at all hard to do with 55 tosses (see screenshot).

You're running these one at a time?

Try this. Go to the "advanced" page. Select binary format. Because we want 55 flips per sample, and 55 factors into 11 and 5, you can generate 11 bits of number by selecting a range from 0 to 2047, and the default 5 columns gives you one sample per row. Generate your maximum of 10,000 numbers per request, and you get 2,000 samples.

With one such sample I analyzed (gvim/cygwin fun) for run lengths, I came up with the following:

[size=1]
1 *
2 ****
3 **
4 **
5 **********
6 **********
7 ***
8 ***
9 *
[/size]

Every "*" represents one sample of 55 flips, or 0.05% of the total (so every 2 is 0.1%). This graph shows all of the runs that max out at run lengths of 3. The numbers are the number of runs of 3 in each sample. The theory behind gambler's fallacy patterns in random data is that people tend to avoid run lengths, so fewer and shorter runs count (run lengths equaling 2 are obviously meaningless--to avoid a run of length 2 you have to alternate, which nobody's going to do when trying to look random).

Your sequence had a whopping 3 runs of length 3--there are seven sequences with 3 or fewer runs of length 3 topping out at 3 in this diagram, representing 0.35% (or 0.0035) of the 2,000 samples. There are 0.45% of the border case just after--where there are four or fewer runs of length 3, topping out at 3. Grand total, there are 36 runs that top out at 3, representing 1.8% of the total.

... Looks more biased than ...

Seriously, you actually believe that?

No, I don't believe that. That's why I didn't say that, and in fact, said quite the opposite (that the latter was more special).

What I did say is that the former is a stronger implication of intent. A two headed coin will give you the all heads flips, without even being sentient. But it's a lot to ask of any weighted coin, no matter how you weight it, to show a strong bias towards GF looking patterns. Ordinary weighting doesn't produce those kinds of patterns.

You will see runs of 3 or less on 55 tosses if you spend just a few minutes generating random numbers.

Or 15 seconds, and a few minutes analyzing the results of the 2,000 samples. As I said previously, with samples merely going up to 55, we're not dealing with insane probabilities here--we're talking on the order of single digit percentages tending towards the decimal end.

You can spend your whole life flipping coins and likely never see 55 heads or tails in a row.

Unless it's a two-headed coin, in which case it's certain you'll see it after the 55th flip. If it's a heavily weighted coin, you might see this too.

But there's no ordinary weighted coin that will show you the GF pattern with such likelihood that the two headed coin will show you all heads. Neither ordinary weighted coins, nor two headed coins, are very sentient devices, so you shouldn't describe their outcomes as being intended.

Now, humans can certainly come up with sequences containing all heads. But it's not insanely hard--and it's not even that much harder at all--for them to come up with sequences showing heavy GF patterns. Humans can even, with enough practice, come up with reasonably good random number sequences.

But both humans and a lot of dumb objects can come up with the most special cases--the all heads case. And both humans and real coins can come up with pretty good random number sequences. But where you find more humans in the set, and less non-sentient alternative causes, is in the data that shows GF patterns.

So if you just invert this, then you see that it's not the case that the more special a thing is, the more likely it is that it was done with intent.

Definitely? This looks pretty random to me. No runs over 3 on 55 "tosses":

Here?

[SIZE=1]2 1 2 2 1
1 2 2 1 1
1 1 2 2 2
2 1 1 2 1
2 2 1 2 1
1 2 1 1 2
2 1 2 2 1
1 2 1 2 1
1 2 1 1 2
1 2 1 2 1
2 1 2 2 2
(TS 2009-09-02 18:37:09 UTC)[/SIZE]

Row 2 has two one's at the end. Row 3 continues with two one's. That's a run of four 1's. Right after that, there are three 2's, and carrying to the next row, another 2. So that's two runs of length 4 back to back.

Another screenshot

...that one has 5 runs of length 3. FYI, it'd be easier to analyze these in text form than JPG--the latter requires either manual calculating, or manual typing/verifying, which is prone to error. In the former, I can just split the numbers to different lines and filter by "uniq -c".

 

[westprog]

The best response to the fine tuning argument I've ever seen is that it's the equivalent of me marvelling at the fact that not one of my ancestors, right back to the first cell, died childless. I mean, what were the odds against that?

The Anthropic Principle is almost unparalleled in its ability to generate inappropriate analogies.

 

[fls]

The best response to the fine tuning argument I've ever seen is that it's the equivalent of me marvelling at the fact that not one of my ancestors, right back to the first cell, died childless. I mean, what were the odds against that?

I guess it depends upon what could be found to interest us in those billions of invisible children (those whose ancestors did die childless).

Linda

 

[fls]

What I did say is that the former is a stronger implication of intent. A two headed coin will give you the all heads flips, without even being sentient. But it's a lot to ask of any weighted coin, no matter how you weight it, to show a strong bias towards GF looking patterns. Ordinary weighting doesn't produce those kinds of patterns.
Or 15 seconds, and a few minutes analyzing the results of the 2,000 samples. As I said previously, with samples merely going up to 55, we're not dealing with insane probabilities here--we're talking on the order of single digit percentages tending towards the decimal end.Unless it's a two-headed coin, in which case it's certain you'll see it after the 55th flip. If it's a heavily weighted coin, you might see this too.

But there's no ordinary weighted coin that will show you the GF pattern with such likelihood that the two headed coin will show you all heads. Neither ordinary weighted coins, nor two headed coins, are very sentient devices, so you shouldn't describe their outcomes as being intended.

Now, humans can certainly come up with sequences containing all heads. But it's not insanely hard--and it's not even that much harder at all--for them to come up with sequences showing heavy GF patterns. Humans can even, with enough practice, come up with reasonably good random number sequences.

But both humans and a lot of dumb objects can come up with the most special cases--the all heads case. And both humans and real coins can come up with pretty good random number sequences. But where you find more humans in the set, and less non-sentient alternative causes, is in the data that shows GF patterns.

So if you just invert this, then you see that it's not the case that the more special a thing is, the more likely it is that it was done with intent.

That's a very nice explanation of the issue (much better than my attempt).

Linda

 

[rocketdodger]

The use of assumptions in forming hypotheses is an essential element of science. That's why we call them assumptions - to distinguish them from claims.

In any case, if the physical constants could not have any other value, that still does not remove the question of why models of the universe need to be so precisely fine-tuned.

You just don't seem to get it.

The concept of fine tuning doesn't even make sense unless there are multiple values to be selected from.

And you -- nor anyone else -- has any idea about the number of possible values each constant could take.

Just because researchers can speculate on what would happen were a constant Y to have value X does not imply that constant Y could actually have a value of X.

I can speculate on what the results of a game of yahtzee might be if a six sided die could take on a value of 5.5. Does that mean a six sided die can actually come up 5.5?

 

[rocketdodger]

The Anthropic Principle is almost unparalleled in its ability to generate inappropriate analogies.

I don't suppose you would care to provide an argument for why that analogy is inappropriate?

 

[Robin]

I just (gasp!) tossed an actual coin 55 times. This is the result:

TTTTHTTTTTHTHHHTHHHTHTHTTTHHTHTTTHHHHHTTTTHTHTTTTHTTTTT

Australian 10 cent coin
THHHHTHTTHTTHTTHTHHTTTHTTTTTHTTTHTHTHHHTTTTTTHTTTTTTTTT

Where would PEAR and GCP be today without those runs?

Now if someone could only cause 55 universes and see how many of them were life supporting we might get somewhere.

 

[fls]

I don't suppose you would care to provide an argument for why that analogy is inappropriate?

I can tell you why I find it inappropriate (don't know if that is why Westprog finds it inappropriate).

We wouldn't find any other way to distinguish between the missed opportunities and the actualized opportunities beforehand (all opportunities result in a human wondering at the improbability of their ancestral chain). The point of considering the universe fine-tuned is that we can independently distinguish between those missed opportunities (universes with different parameter values) and actualized opportunities (our universe) beforehand by looking at what sort of universe will result. The question is whether or not there are other ways to sort universes than by 'maximize the variety of objects'.

Linda

 

[Robin]

No, it's based on the fact that you maxed out at 3, and that you only had 3 runs of 3 or more.

You're running these one at a time?

Try this. Go to the "advanced" page. Select binary format. Because we want 55 flips per sample, and 55 factors into 11 and 5, you can generate 11 bits of number by selecting a range from 0 to 2047, and the default 5 columns gives you one sample per row. Generate your maximum of 10,000 numbers per request, and you get 2,000 samples.

With one such sample I analyzed (gvim/cygwin fun) for run lengths, I came up with the following:

[size=1]
1 *
2 ****
3 **
4 **
5 **********
6 **********
7 ***
8 ***
9 *
[/size]

Every "*" represents one sample of 55 flips, or 0.05% of the total (so every 2 is 0.1%). This graph shows all of the runs that max out at run lengths of 3. The numbers are the number of runs of 3 in each sample. The theory behind gambler's fallacy patterns in random data is that people tend to avoid run lengths, so fewer and shorter runs count (run lengths equaling 2 are obviously meaningless--to avoid a run of length 2 you have to alternate, which nobody's going to do when trying to look random).

Your sequence had a whopping 3 runs of length 3--there are seven sequences with 3 or fewer runs of length 3 topping out at 3 in this diagram, representing 0.35% (or 0.0035) of the 2,000 samples. There are 0.45% of the border case just after--where there are four or fewer runs of length 3, topping out at 3. Grand total, there are 36 runs that top out at 3, representing 1.8% of the total.


No, I don't believe that. That's why I didn't say that, and in fact, said quite the opposite (that the latter was more special).

What I did say is that the former is a stronger implication of intent. A two headed coin will give you the all heads flips, without even being sentient. But it's a lot to ask of any weighted coin, no matter how you weight it, to show a strong bias towards GF looking patterns. Ordinary weighting doesn't produce those kinds of patterns.
Or 15 seconds, and a few minutes analyzing the results of the 2,000 samples. As I said previously, with samples merely going up to 55, we're not dealing with insane probabilities here--we're talking on the order of single digit percentages tending towards the decimal end.Unless it's a two-headed coin, in which case it's certain you'll see it after the 55th flip. If it's a heavily weighted coin, you might see this too.

But there's no ordinary weighted coin that will show you the GF pattern with such likelihood that the two headed coin will show you all heads. Neither ordinary weighted coins, nor two headed coins, are very sentient devices, so you shouldn't describe their outcomes as being intended.

Now, humans can certainly come up with sequences containing all heads. But it's not insanely hard--and it's not even that much harder at all--for them to come up with sequences showing heavy GF patterns. Humans can even, with enough practice, come up with reasonably good random number sequences.

But both humans and a lot of dumb objects can come up with the most special cases--the all heads case. And both humans and real coins can come up with pretty good random number sequences. But where you find more humans in the set, and less non-sentient alternative causes, is in the data that shows GF patterns.

So if you just invert this, then you see that it's not the case that the more special a thing is, the more likely it is that it was done with intent.

Here?
[qimg]http://www.internationalskeptics.com/forums/imagehosting/thum_268924a9f1df8ee9cb.jpg[/qimg]

[SIZE=1]2 1 2 2 1
1 2 2 1 1
1 1 2 2 2
2 1 1 2 1
2 2 1 2 1
1 2 1 1 2
2 1 2 2 1
1 2 1 2 1
1 2 1 1 2
1 2 1 2 1
2 1 2 2 2
(TS 2009-09-02 18:37:09 UTC)[/SIZE]

Row 2 has two one's at the end. Row 3 continues with two one's. That's a run of four 1's. Right after that, there are three 2's, and carrying to the next row, another 2. So that's two runs of length 4 back to back.

...that one has 5 runs of length 3. FYI, it'd be easier to analyze these in text form than JPG--the latter requires either manual calculating, or manual typing/verifying, which is prone to error. In the former, I can just split the numbers to different lines and filter by "uniq -c".

Nominated

 

[westprog]

I can tell you why I find it inappropriate (don't know if that is why Westprog finds it inappropriate).

We wouldn't find any other way to distinguish between the missed opportunities and the actualized opportunities beforehand (all opportunities result in a human wondering at the improbability of their ancestral chain). The point of considering the universe fine-tuned is that we can independently distinguish between those missed opportunities (universes with different parameter values) and actualized opportunities (our universe) beforehand by looking at what sort of universe will result. The question is whether or not there are other ways to sort universes than by 'maximize the variety of objects'.

Linda

I think that's a fair summary.

If we have a bag full of a trillion numbered marbles, then the odds are against any particular number coming up. So to be astonished when we pull out a particular number at trillion to one odds is silly. However, if the bag has a trillion white marbles, a trillion black marbles, and one red marble, and we pull out the red marble on our one and only try, we can be astonished, and wonder how it happened.

One might claim that we only consider this universe special, out of all the other universes which we have modelled, just because we live in it. But if we can say, objectively, using reasonable criteria, that this universe lies in a very small subset of the different types that we model, then we are, I think, entitled to delve a bit deeper.

 

[westprog]

You just don't seem to get it.

The concept of fine tuning doesn't even make sense unless there are multiple values to be selected from.

And you -- nor anyone else -- has any idea about the number of possible values each constant could take.

Just because researchers can speculate on what would happen were a constant Y to have value X does not imply that constant Y could actually have a value of X.

Since you are rephrasing things that I've already said, and serving them back to me, they don't seem very cogent objections.

Either the values could, in some sense, have had different values, or they couldn't. In either case, it's a matter of interest why they have the values they do.

I can speculate on what the results of a game of yahtzee might be if a six sided die could take on a value of 5.5. Does that mean a six sided die can actually come up 5.5?

You are picking an example that you know to be impossible. You don't know that it's impossible for physical constants to have alternative values.

I find it strange that you are objecting to investigation of such matters, and at the same time when I point this out, you deny it. Should we be considering why physical constants have the values they do, or not?

 

[rocketdodger]

I can tell you why I find it inappropriate (don't know if that is why Westprog finds it inappropriate).

We wouldn't find any other way to distinguish between the missed opportunities and the actualized opportunities beforehand (all opportunities result in a human wondering at the improbability of their ancestral chain). The point of considering the universe fine-tuned is that we can independently distinguish between those missed opportunities (universes with different parameter values) and actualized opportunities (our universe) beforehand by looking at what sort of universe will result. The question is whether or not there are other ways to sort universes than by 'maximize the variety of objects'.

Linda

But that difference only exists if you lack imagination.

The difference between myself and the human that I might have been is slight. The difference between a universe like ours that contains life as we know it and some other universe that contains something else that might wonder about it's existence is grand.

Yet, the two are fundamentally the same type of difference. One is me, the other is not. A totally different universe in no way implies the lack of an entity that could reason about it's existence.

 

[fls]

But that difference only exists if you lack imagination.

The difference between myself and the human that I might have been is slight. The difference between a universe like ours that contains life as we know it and some other universe that contains something else that might wonder about it's existence is grand.

Yet, the two are fundamentally the same type of difference. One is me, the other is not. A totally different universe in no way implies the lack of an entity that could reason about it's existence.

I agree. The fine-tuners are pretending that the bag is full of marbles with only one or a few differing in colour. In reality, the bag is full of a million marbles of different sizes and shapes, half a dozen kittens, a few hundred socks, several sets of house keys, an umbrella, 60,000 glass beads, 20 balls of crumpled paper and a used condom. They are wondering why a kitten was pulled out, considering that there were only a few grey things in the bag.

Linda

 

[rocketdodger]

Since you are rephrasing things that I've already said, and serving them back to me, they don't seem very cogent objections.

Either the values could, in some sense, have had different values, or they couldn't. In either case, it's a matter of interest why they have the values they do.

But that is not what I am objecting to, and I think you know it.

I am objecting to understanding the above and then proceeding to act otherwise while making an argument. And again I cite Malerin, who knows darn well that there is no data to support the notion that the constants could, for instance, take on any real number, yet proceeds to generate a likelihood of the constants being what they are close to zero as if they could take on any real number.

And you seem to be in the same boat. Do you or do you not support the aspect of the FT argument that claims the likelihood of the constants being what they are is low?

You are picking an example that you know to be impossible. You don't know that it's impossible for physical constants to have alternative values.

If I didn't know about dice, I wouldn't know it was impossible. If all I saw was a single result of a die roll, and someone told me "some process" generated the number, what should I think?

I really would like to hear your answer on that. If an unknown process generates a value, what can you conclude?

I find it strange that you are objecting to investigation of such matters, and at the same time when I point this out, you deny it. Should we be considering why physical constants have the values they do, or not?

I am not objecting to the investigation. I am objecting to the use of results that do not exist since the investigation hasn't even begun.

Of course we should be considering why the constants have the values they do. Do you think I am stupid? What we should not be doing is exactly what Malerin and others are doing, which is assuming the likelihood of any values of the constants.

 

[westprog]

But that difference only exists if you lack imagination.

The difference between myself and the human that I might have been is slight. The difference between a universe like ours that contains life as we know it and some other universe that contains something else that might wonder about it's existence is grand.

Yet, the two are fundamentally the same type of difference. One is me, the other is not. A totally different universe in no way implies the lack of an entity that could reason about it's existence.

Personally, I think that it's quite unlikely that a universe that collapses on itself after an infinitesimal fraction of a second will generate something that will wonder about itself. Nor do I think that a family of such universes, differing only by the tiny difference in time before they collapse, would objectively be considered as different from each other as they are from our universe.

Of course it's difficult to come up with objective standards for classifying universes that only exist in physical/mathematical models, but it's certainly not impossible. To me, many of the hypothetical universes look like black or white marbles, not kittens or bananas.

 

[westprog]

What we should not be doing is exactly what Malerin and others are doing

I suggest you address your concerns to them.

 

[rocketdodger]

I agree. The fine-tuners are pretending that the bag is full of marbles with only one or a few differing in colour. In reality, the bag is full of a million marbles of different sizes and shapes, half a dozen kittens, a few hundred socks, several sets of house keys, an umbrella, 60,000 glass beads, 20 balls of crumpled paper and a used condom. They are wondering why a kitten was pulled out, considering that there were only a few grey things in the bag.

Linda

Oh, so thats what you meant by asking whether there are other ways to sort universes?

 

[westprog]

And you seem to be in the same boat. Do you or do you not support the aspect of the FT argument that claims the likelihood of the constants being what they are is low?

I think that Beth expressed this quite well. When modelling the universe, cosmologists have found that if the values of the constants are not expressed to an astonishing degree of precision, the model doesn't work as a representation of what actually happens. That is the strange phenomenon that we're dealing with.

 

[fls]

Oh, so thats what you meant by asking whether there are other ways to sort universes?

Yes.

Linda