Continuing the discussion from Eric Holloway: Algorithmic Specified Complexity:

It is commonly asserted that CSI, a type of mutual information, is a unique signature of intelligence, and that the CSI content of any object can be easily measured with metrics ASC. The argument continues, we observe CSI in life, so therefore it must be intelligently designed. As @EricMH, one of Robert Marks’s PhD students, explains…

So I have a puzzle for ID advocates to solve that seems impossible to me. I know the answer, just because I constructed the strings. Who can can solve this puzzle? If any ID advocate can, I will be truly impressed, and have much to learn from them.

1 Like

Could you explain why this follows?

I should rephrase:

If you can’t solve this puzzle with an objective metric, it means you can’t objectively measure the CSI of arbitrary objects.

If that is the case, there is no way to objectively demonstrate the unique signature of minds on arbitrary objects, such as DNA. Of course, as I have explained elsewhere, MI is not unique signature of minds any ways.

This is fallacious reasoning. Just because we cannot measure ASC in some instance, i.e. encrypted text, does not mean we cannot measure ASC in some other instance.

1 Like

Correct! I’ve always agreed with this.

However, Dembski, Durston, and Marks have been arguing they can use ASC on DNA to quantify FI, ignoring how it was produced, because they argue that ASC is always an underestimate of the true CSI. We’ve already clarified that this only true if they use the right P. Durston, however, does not even attempt to improve upon a uniform distribution. It is clearly the the wrong P, and therefore overestimates CSI. They were hoping to avoid this critique by ignoring the impact of picking the wrong P altogether.

So, this is a substantial set back for their current best argument.

So now, if can continue to identify example after example where their metric fails, we can understand what characteristics cause problems for them. You hit on one of them with “encryption”. DNA is much more like an encrypted program than a clear text program. There is more though. That is not the whole story. We will several layers of qualities that DNA have that frustrate application of ASC like metrics to determine the difficulty of evolving something.

In the end, we can still measure ASC in carefully selected toy examples. The key question is if it works in real world biology. The honest answer is not yet, and maybe not ever.

1 Like

A uniform distribution is not clearly wrong. Especially if DNA cannot be generated altogether by naturalism and thus has a probability of zero.

And even if it is wrong, that doesn’t invalidate what they are doing. It just means they have more work to do. It also means that ID makes empirically workable claims that could be verified or falsified, and thus is a science.

Go understand the cancer thread, and see then if you understand why it is clearly wrong.

@swamidass, wow!

@sfmatheson, have you seen this challenge yet? Do you think it is solvable?

I faintly recall the challenge. If I understand correctly, you have some strings that contain embedded “specified information.” I know nothing of encryption/decryption except that both exist and both presumably depend on principles of information and pattern and so on. So, it seems to me that a method that can detect “CSI” in biological polymers should be able to detect it in encrypted text. If that method can’t work on your strings then it is not really a method to detect “CSI” and is more likely a fancy post hoc construction to get the authors to a highly desired apologetic endpoint. Am I close?

1 Like

encryption adds information to the original message that is only known to the sender and receiver. The receiver uses that information to decrypt the original message.

1 Like

Does that mean a CSI detector can’t work in such a situation?

The process added and scrambled information, but I did not explain how. To give a little more information now, I used an encryption filter that allows me to precisely set the CSI of each string. This is just a few lines of python code I stored away for a later date.

Because we don’t know the process I used, there is no way to compute the CSI.

I randomized a number in the acceptable zone from a larger domain. That acceptable zone size is tuned to equal the CSI. Then I added a secret key and “shuffled” the bits. That’s the process.

Now you know the process, however you still can’t compute the CSI, because I didn’t tell you the secret key, or the precise domain and acceptable zones I used, or how I shuffled the bits around. Even I couldn’t tell you which one was which unless I had kept notes to tell which one is which. (I did)

From here it should be clear that it would require cheating or mind reading to solve this task. Unless we know the process that constructed a sequence, and unless we know the specification, we cannot actually, compute the CSI.

2 Likes