Signals vs. Noise

Recently an Evolution 2.0 reader reached out. He’d done deep study of the linguistics of DNA.

Because I’d sine_fernando_marcelino mentioned designing speakers and stereo equipment in chapter 3 of my book, he sent a photo of a some speakers he’d built.

He had no idea I also built a system that was spooky similar. Same design idea. Same ribbon drivers from the same boutique German manufacturer.

I sent him back some pictures. We both got a chuckle from this “needles in haystacks” meeting in such an unlikely context.

The next week another guy reached out to him, wanting to discuss nuances of DNA’s digital code. Turns out that guy was into something similar. I think it was Digital Signal Processing.

Then it struck me:

What all three of us have in common is deep hands-on experience of how FRAGILE signals are. This is why we’re so fascinated with DNA’s finely layered data structure.

In audio, once you’re damaged a signal, there’s no going back. Once you’ve lost information, you can never recover it. Period. If you copy a master tape onto an LP then destroy the master, there’s no getting back to the original. (Let alone getting something better.)

You work like crazy to preserve the signal’s integrity every step of the way. The signal you hear from a speaker is still never as good as what went into the microphone. 

When you work in the audio profession the way I did, you know IN YOUR BONES how fragile that data is. And data is data. The guy who manages databases or installs fiber optic cables has the same intimate knowledge.

This is why people with our background know IN OUR BONES that copying errors would never make DNA better. Ever. “Accidental mutations” are garbage. Garbage is garbage. All the natural selection in the world can’t fix it.

We know this.

This is why nobody who truly understands signals and noise buys into old-school Darwinism. Not if they understand what the Darwinists are really saying, anyway.

This is how I knew, very early in my search for the truth about evolution, that most of the biology profession had missed something very VERY big.

Some people believe in the Easter Bunny.

Some people believe in the Tooth Fairy.

Some people believe DNA copying errors help life evolve.

The evidence for all of these things is zero.

This is why I wasn’t the least bit surprised when I found out cells employ elaborate machinery for detecting and repairing errors. I wasn’t surprised to find out the the failure of this error correction is often the cause of cancer.

I wasn’t surprised to find out this is a HOT field of research – including a 2015 Nobel Prize. It’s a shame more people don’t know about it.

This is why when I discovered cells cut, splice, edit and re-structure their own DNA – in response to what’s going on around – I realized one of the great secrets in the history of science had been hiding right under our noses. In plain sight.

And most people had walked right on by.

2 Responses

  1. John Chalisque says:

    This is a general reply to a number of points made on your site:

    In response to the email DMZ part 3:

    Science is all about what you can’t prove wrong. It is very counter intuitive. A theory is only as good as the degree to which it has been experimentally tested. The point of a scientific theory is to give an explanatory theory for why things are the way we observe, and hopefully offer the ability to make real world predictions about what we have not observed yet, but could in the near future. It is like the work of a stone sculptor, beginning with all possible theories, and chiselling away those which do not fit the evidence. Just as a stone sculptor, sculpting a rhino, begins with an uncarved block, and slowly and carefully removes all that is not a rhino, so the scientist begins with all possible explanations and removes all those which do not fit the evidence. A big problem with science, both within the discipline and especially in describing it to people in general, is not grasping this ‘topsy turvy’ nature: whilst you are ultimately interested in what is true, it is by focussing your attention on hunting for reasons why you’re wrong, that marks the discipline out from more naive quests for truth. A practical engineer knows that the test of cable is what it takes to break it; and a famous person brought up by a carpenter and is wife a couple of millennia ago used a similar analogy: that of a house’s stability in the face of heavy weather is only as good as its foundations. If people are not trying to prove themselves wrong, but are taking what they believe or want to be true and pushing it as hard as they can, it is not science, but something else. Richard Feynmann, in his talk Cargo Cult Science, gave a very stern warning about the rise of ‘science that is not science’. So far as forming common ground between people involved in this debate, Richard Feynmann’s popular writings (the meaning of it all, for one thing) are a good place to start. Just writing out your position and hunting for a number of distinguished talking heads to say stuff in your favour is not enough, nor is it helpful. Just about everybody can find plenty of distinguished talking heads to endorse their writings, and the only use for such endorsements is in working out whose endorsements are not worth listening to.

    With regards to the ‘new theory of evolution’
    1. Your claim ‘Nowhere in the vast field of engineering is there any such thing as “the percentage of the time that corrupted data is helpful instead of harmful.”’ is simply not true. The most obvious is in audio production when downsampling to a lower bitrate, where deliberate introduction of noise mitages the problems of quantization. By introducing a small amount of noise, and shaping it, the resulting signal can be improved. Likewise, with evolution, introducing small errors and then shaping the result, can result in improvements. The idea that problems are ‘ALWAYS harmful’ is something which, if true, would require rigorous substantiation, in the form of mathematical proofs that introducing problems can never produce beneficial outcomes. The reality in evolutionary computing shows well the problem of local optima: where every _very small_ change from a particular point on the evolutionary landscape is necessarily downhill (that is, worse). But from a local optimum which is not globally optimal, the only way to begin a path to a better optimum is downhill, that is, you have to get worse at first, then get better again, but in a different direction.

    With regards to scientific theories in general:
    They are always oversimplifications, ignoring myriad details. A good theory, in terms of real world predictive power, ignores only what is negligible. Many theories do not do this, but gain much simplicity and are good for illustration, but not for actual real world prediction. Simple Darwinian Evolution is a good means to give the basic idea of how things evolve to naive learners, but what must be stressed is that the reality is much more complex, and the behaviours possible by simple Darwinian Evolution are already so massive that we can’t even fully understand them.

    In general, though, with science the most important thing is to have a common ground in terms of _what we are trying to achieve_ with a particular theory, and then address the question of whether the theory achieves that intended goal, and whether it achieves any other goals.

    Finally, with regards to the DMZ issue, the basic need is for science to be de-politicsed, de-religionised and de-commercialised. Unless scientific progress only, or primarily answers to the need for a detailed reliable understanding of how the world works, and does not pander to the interests of any particular party, science will degenerate into a chaotic miasma of marketing, religion and politics. It has already come quite far in that journey.

    • 1. Your claim ‘Nowhere in the vast field of engineering is there any such thing as “the percentage of the time that corrupted data is helpful instead of harmful.”’ is simply not true. The most obvious is in audio production when downsampling to a lower bitrate, where deliberate introduction of noise mitages the problems of quantization.

      Yes that is called dither and I discuss that in my book Evolution 2.0 in Appendix 1. The noise is applied very deliberately and it doesn’t add information to the signal.

Leave a Reply

Your email address will not be published. Required fields are marked *