Don't believe the hype from the media about science

n my most recent paper, I have a big paragraph talking about how beta-amyloid is involved in Alzheimer's disease. If someone counted up papers, mine would be counted as one of the 99% that say beta-amyloid causes Alzheimer's disease.

But I do not believe that. The peer-reviewers explicitly demanded that I add the paragraph. If I hadn't, it would not have been published.

On my computer, I have 972 reprints of articles claiming that Alzheimer's is a disease. If I counted those, I'd find that more than 99.89% of scientists believe it is a disease and not a natural part of aging. Does that mean there's a consensus and it is “case closed”?

It does not. It could be a disease or it could be something else. It simply means that the US govern­ment has a near-total monopoly on research funding, and the US government wants it to be a disease, and so you have to say it's a disease or you won't get funded. Besides, how would you prove that it's not? I can't speak for those who consider it not to be a disease, but I imagine it's more a question of why bother. Every scientist knows it's nearly impossible to get a negative result published.

You'd never know it by reading the editorial page at Nature and Science, but the libertarian spirit is still strong in science, even among those who vote straight D. There's no percent­age in shooting down someone else's theory. Most scientists just want to be left alone and discover stuff. In return they'll try not to blow up the world—but no promises.

Global warming

Now that it's late September and global warming is over again, we find the news media praising a 2021 article ^[1] by some guy named Mark Lynas who says he read 2718 articles discussing “global warming” or “climate change”. Miraculously, he survived that ordeal and found, according to his abstract, only four that were skeptical of global warming. The authors conclude that the widely debunked 97% figure was way too low:

Our finding is that the broadly-defined scientific consensus likely far exceeds 99% regarding the role of anthropogenic GHG emissions in modern climate change, and may even be as high as 99.9%.

However, if one looks at their pie chart, it tells quite a different story. Their actual results show that there is nowhere near the claimed 99.85% consensus:

Depending on what is meant by “implicit endorsement,” what the actual data show is that at most 31.09% of the papers endorsed the AGW theory and 69.58% did not: a greater than 2:1 ratio of no-position + rejection over endorsement. This is what I would call lukewarm support at best, considering that the funding and professional advancement of these researchers depend on supporting the AGW narrative. Discussing AGW while taking no position on its validity is a defensible course for skeptics in the field who wish to remain employed.

This lack of support for the global warming narrative by two-thirds of the very people whose careers depend on it is nothing short of astounding. Perhaps that is why Lynas et al. so badly misin­terp­reted their own results. They threw away their chance to blow a previous claim out of the water and bring back honest debate on this topic. Sadly, altering one's conclusions to conform to a pre-established view happens frequently in science.

There's another aspect of this. The appearance of false statements and exaggerations in a scientific field is hardly unprece­den­ted. It's an unmistakable sign that the field is headed for a downfall and that its adherents are, as my grandmother would say, hanging on by the skin of their teeth. They sense the impending doom and raise their tenor accordingly.

As many have pointed out, consensus doesn't mean squat in science. We all know you can't get a climate study funded unless you relate it to global warming. AGW is their cash cow. The government wants it to be true, and the scientists who rely on government funding are in a tough spot. Are there papers that say AGW is an established fact? Yes. But whether they really believe it or not is another matter.

Artificial intelligence

AI is another field that's showing the same symptoms. There's a press report of a paper claiming that AI is 90% accurate at deciding whether there is extraterrestrial life. This comes from a PNAS article by H. James Cleaves II et al. [paywalled, so link only] that used pyrolysis-GCMS, a method that heats samples to decomposition and then measures the products that are produced. They trained the AI on a variety of samples from planet Earth, including degraded ancient biological samples, to create a fingerprint. The authors say actual identification of the molecules doesn't matter, but the program could predict if the sample had a biological origin with 90% accuracy.

Gas chromatogram from an unrelated study on a 100-meter-long GC column. This is a typical GC tracing from a sample found somewhere on planet Earth. It shows at least 917 different molecules. Pyrolysis would break each one down into fragments and show many more. According to Hübschmann [2], pyrolysis produces an extremely large number of reaction products that cannot be completely separated even under the best GC conditions

I've seen claims like this many times, usually in the context of disease diagnosis. Somebody runs a sample through a mass spectrometer. The mass spec gives you some huge number of peaks, each representing a different molecule or fragment of a molecule. You run a big correlation analysis and conclude, lo and behold, some combination of peaks predicts the disease with 100% accuracy.

Unfortunately, such studies almost never hold up. One reason is that for many diseases a pathologist can only get an accurate diagnosis between 50 and 80% of the time (which is why everybody needed a biomarker for it). A pathology report is the gold standard against which their results are meas­ured, so a marker cannot even in principle be more accurate than the pathologist. Another reason is that in most cases we have no clear idea what the statistical analysis is actually ‘statistic-ing’ on. So whatever their mass spec is measuring, there's a good chance it's a statistical artifact.

With regard to the alien life claim, a moment's reflection should tell you that we have absolutely no clue what molecules it may contain, so all the GCMS test tells us is whether the sample has been contaminated by Earth molecules. To get a valid statistic, they would at a minimum have to have a control sample from extraterrestrial life and another control from extraterrestrial non-life. The author, of course, realizes this and says “[W]e cannot assume that alien life will use DNA [or] amino acids.”

What he means is that the hype has far outrun the reality and he and the other authors know it, but by gum they're still going to cash in while the AI fad lasts. When a branch of science is overrun with hype and fake results, it's a sign that it's about to collapse. That's bad news for all those people buying stock in artificial intelligence—and global warming.

sep 27 2023, 4:31 am. last updated 3:33 pm