Thursday, October 21, 2021

Why "science replaced by hodgepodge" is even worse than alchemy



Maritime hodge podge

For two days, I wanted to write another text about the evil trend of "pretending that some chaotic playing with computers is the new science" and Robbert Dijkgraaf, the director of the IAS, pushed me above the critical threshold when he wrote
The Uselessness of Useful Knowledge
in the Quanta Magazine. The title looks bizarre, the normal thing to describe the body of his essay would be the opposite, "the usefulness of useless knowledge", a title that was picked by several authors in the past. Isn't it a typo? At any rate, Dijkgraaf argues that the "hocus pocus AI-based quasi-science" is analogous to alchemy which may look bad but it was very useful. And the periods of chaotic expansion of science and the epochs of systematic consolidation are equally important and they alternate which was good and it is good now, too.



Note that his article takes the approximately opposite position than my
Worshiping of complexity is junk physics, junk computer science, junk neuroscience
that I wrote six weeks ago. Also, I have some feeling that Dijkgraaf decided to become an apologist for the totally atrocious choice of the 2021 Nobel Prize in Physics that ended the epoch of credibility of that award among sensible people.



OK, is the Artificial Intelligence-exploiting hocus pocus analogous to alchemy? Yes, indeed, I think it is a very good comparison. The people doing both weren't or aren't following the scientific method in the strict, Galilean sense. They aren't really finding any laws that really work. They don't impartially choose the best theory explaining some phenomena; they're not abandoning theories that have contradicted the data; they don't care about the range of validity of their claims; they don't feel the duty to mention the error margins of most numbers in their propositions, or most of the uncertainties of many kinds.

Most importantly, both alchemists and the hocus pocus AI researchers generally had or have a very muddy relationship to the truth itself – and they often conflated or conflate the "truth" with their "ability to survive or make a living". These are extremely different things for anyone who actually has the moral qualities to be a scientist. When an alchemist made the emperor Rudolph II believe that he was close to an elixir of youth (click: in that article, I discussed that Unity Biotechnologies indeed restarted alchemy with all of its antiscientific misinterpretations of related concepts, phenomena, and methods), and Rudolph II increased the alchemist's salary, this alchemist considered it a success even though the alchemist knew that his promise was almost certainly a lie. This kind of corruption and the lack of integrity is widespread in the computer hocus pocus research, too.



Was alchemy a good epoch in the evolution of science? First, alchemy wasn't a science in the strict sense. Dijkgraaf calls it a "protoscience" (an activity that is not fully scientific but resembles science in many ways; and that often preceded genuine science) and "hyperscience" (because the people overpromised things, like the elixir of youth or the production of gold from cheap materials). We may surely find useful things that were found by the alchemists, sort of accidentally. But the amount of resources that went into alchemy was immense and the "products" just didn't have this value. Note that alchemy became a big industry in the Greco-Roman Egypt shortly after Jesus Christ's life. And as I mentioned, the reign of Rudolph II in Prague (around 1600) was still a golden epoch of alchemy (or at least a urine epoch whose color was claimed to resemble gold). For well over 1,000 years, a big part of numerous rulers' money went to the alchemy research. What have they found?

They were promising to cheaply produce gold, to cure all diseases, and to guarantee immortality to the consumers. None of these things has resulted from that research of alchemy. And I think that modern science really implies that it couldn't have. If these things are possible, they must be done differently than by the naive hocus pocus that the alchemists were repeating for over 1,000 years. Nuclear reactions simply aren't in one-to-one correspondence with a sequence in which carrots and other vegetables are being added to a hodge podge. There is no permutation of vegetables that produces gold! One may argue that this result was unavoidable. A real scientist (and even an honest person), before he starts to refine the details of the pledge to create gold out of šit, should impartially ask the good question whether it's possible to turn šit into gold at all! Before any successful research is done, both YES and NO must be considered possible correct answers (in his Cargo Cult Science speech, Feynman also discussed why it's scientifically dishonest to publish results of a certain kind – pushing things in a certain direction – only). The transmutation is possible but only through nuclear reactions that are much more modern processes (and they demand more powerful gadgets) than the inefinitely repeated cooking by the alchemist kooks. But they didn't care about the answers that were not useful for the increasing of their salaries and the salaries resulted purely from the hypothetical "supernaturally useful products". Good science can't really flourish in these skewed and twisted conditions. Not only the "pure theory" suffers when the researchers don't have the scientific integrity; the "applied science" suffers, too (at least with some delay). The scientific integrity is needed; you may view it as something ethical at the level of emotions about morality and outrage; but you may also be cold (I am not) and imagine that the scientific integrity is just a trait or even "another recent invention" that is needed for valuable scientific results to systematically emerge.

Maybe the alchemists did find some chemical processes or reactions or materials. But they didn't really appreciate them because the "right, desirable results" were predetermined. They didn't want to accumulate truths as they happened to arrive; they only wanted to find justifications for predetermined "truths" (which were falsehoods). From the perspective of science, someone else – whose thinking was much more honest, scientific, and rational – had to go through the alchemists' hocus pocus and find something useful over there. Because this was needed, we face a legitimate question who should get the credit for the scientific discoveries. I think that the "later scientist who reviewed all the hocus pocus" should get almost all the credit. It's just like the monkeys that randomly type something. All the possible books are out there. But it's Shakespeare who found the "right books" written by the monkeys that deserves the credit, not the monkeys. Because the hocus pocus researchers are generally unable to separate weed from chaff, their accidentally valuable discoveries should be considered as random natural events, not scientific discoveries. Yes, by doing lots of hocus pocus, you may increase the probability that you find something important. But if your having found something valuable is due to sheer luck and many people who are almost equivalent to you did the same and failed, you are not the "boss" in the activity that matters; it is the person who filters all the random stuff you have found!



So alchemy wasn't great from the viewpoint of the cost-and-benefit analysis. When people started to do chemistry in the true scientific, systematic, honest way, the progress became much faster and cheaper. The neural network hocus pocus is analogous to alchemy. But it's worse than alchemy for one reason. Alchemy was both a protoscience and a hyperscience but it was also
the most scientific kind of the research of chemical reactions that mankind of those times possessed.
This is the fundamental difference between alchemy and the neural network hocus pocus that makes the neural network hocus pocus even worse. The artificial intelligence simulations just aren't the best scientific way to look at the real world biological neuron, the brain, even the climate, and many other things that are being mindlessly "thrown to computer programs including neural networks" these days. We already have systematic scientific disciplines that study all these phenomena and they simply must remain in control of things.

As the Enyaq ad asked, when did we lose it? The fun, the ease, the lightheartedness. The things that make life life and that make science science. I mean, look at us. It's all about techno gadgets, unhinged woke activists, mumbo jumbo, sci-fi nonsense, hocus pocus, the nerd stuff, the future (yes, that's the reference to the insane promises that are being linked to the AI pseudoscience, too: with his dumb projects, e.g. Musk is a perfect contemporary clone of the alchemists). Come on, where's the fun?

Don't get me wrong. The artificial intelligence is already useful and will be even more useful. Try the DeepL.com German neural machine translator. It has become a vastly more professional translator than I could ever be. Try any pair of languages; 99% of the resulting sentences are immediately usable. Try to use Google Lens, an image recognition technology, or the Voice Assistants on your phone. If you don't use them, you will be amazed how far this has gotten. No doubt about it, this practical progress will continue. The practical usefulness of the AI brute force is the only advantage of the AI relatively to alchemy. Brute force used to be much less useful in alchemy. A 100 times greater bowl for hodge podge didn't get much closer to an elixir of youth; a 100 times more powerful computer may bring you closer to a perfect translator.

But the mere usage of similar neural machine technologies (and perhaps the usage of computers in general) is simply not science. It is primarily the "human factor" that is deteriorating. The usage of some computer software, and it's increasingly the neural machines, has become a universal excuse for the people not to understand any science themselves. They think that possessing some neural machine mumbo jumbo replaces the scientific understanding. Sorry, it doesn't and it cannot. Science is about laws that are concise enough to be actually remembered and understood by humans. And while your computer may do many things by brute force, it is not really you who should get the credit if you just pressed the enter key. And if the computer doesn't get the direction how to do something that is genuinely valuable science, it just won't do it, at least not systematically. So far, comments made by AI neural machines are just preprocessed messages picked by a coach. That directly follows from the fact that the artificial intelligence entities don't really live free lives. For them to become "scientists", they need to gain the "personal freedom" first! And that would have many other consequences. At the end, all machine learning is just a multi-dimensional interpolation but science is much more than just an interpolation. It often requires extrapolation or thinking outside the box – and that can only be done by free subjects that may actually leave the box (or their cage or Australia).

Dijkgraaf also discusses other episodes in the history of science that he believes to be analogous to alchemy and to the neural machine mumbo jumbo. A representative slogan says:
Out of the thick mist of practical considerations, mathematical laws slowly emerged.
Two of his examples are the emergence of the laws of thermodynamics from the practical playing with the thermal engines; and the slow emergence of general relativity. Was this progress a slow result of the mumbo jumbo? Yes and no. The steam engine was built in the late 18th century and many people started to play with variations of similar gadgets. In April, I discussed the fact that James Watt had quite some mathematical and physical family background. More generally, people who made genuine "conceptual" progress in sciences usually had some science-like background. Lots of other people with a more ordinary background were only playing and they still made contributions that were important practically.

But one can still divide the people to those who tried to do a serious research of the thermal phenomena; and those that didn't. The latter group was applying the mechanical laws as pioneered by Isaac Newton (and it was enough for them to understand them superficially) while the excess heat was either an unavoidable byproduct of the machines' work, or it was added in a mumbo jumbo way. You should understand that all the transformation of forces that originate in the steam pressure was perfectly understood. It was really just the behavior of the heat and temperatures that wasn't described by reliable equations for some time. But the inventors had some intuition for the behavior of heat from their everyday life.

At some moment in the first half of the 19th century, theorists started to understand the things. Some of the thermal engine insights must have been useful although I am confident that the theory could have been built (and probably was built) using the experimental data from a tiny fraction of the gadgets that had been built. Most of the mumbo jumbo was useless for the progress, much like in the case of alchemy. Yes, it is totally right when the practical mumbo jumbo and the careful theoretical work co-exist. But that is exactly what is under attack now because the defenders of the neural machine hocus pocus want to return mankind back and replace science but the protoscience and chaotic experimentation (that looks like a black box to the humans) again. You might think that they pretend to root for the co-existence but it is as myopic a view as in the case of the SJWs. The termination of meritocracy, in this case of the principle that "the scientifically better methodology must get more attention", is just the first step towards the complete dominance of the inferior methodology. Similarly, the political correctness towards non-white races was unavoidably just the first step towards the systematic suppression of the whites. There are tons of such examples. Societies just mustn't allow the first step in this pernicious line! The people who promoted the political correctness should have been executed already decades ago, it was already too dangerous then. If you don't get this point by now, you have learned nothing from the recent history at all.

Another analogy of Dijkgraaf's is Einstein's discovery of general relativity which also slowly emerged from the mist. So far so good. Einstein has struggled with the general relativity for a decade (a reason it took so long was that he wasn't a real hard worker). And he has described this 1905-1915 decade using the word "mist" with the "light at the end of the tunnel", too. But there are still big problems with Dijkgraaf's spinning of the rise of general relativity. By "slow", he really means "gradual" (the latter adjective appears in Dijkgraaf's article in another analogy that is also meant convey the same message).

While the discovering of general relativity was slow, it wasn't really gradual. It was a lot of random oscillation of Einstein's mind that didn't lead anywhere interrupted by extremely discontinuous events, the real breakthroughs. For example, he realized that the equivalence principle implied the bending of light at a very specific moment in 1913. It was several seconds when he got this "eureka moment". He was in his office in Viničná [Vineyard] Street in Prague and was just looking at several female patients in the psychiatric asylum across the street. We may actually reconstruct the timing and events that took place around that moment rather accurately. Because general relativity is made out of a bunch of very powerful yet concise ideas, the history was analogously made during several key seconds during the decade. It is just BS when Dijkgraaf writes that the whole decade mattered equally. It is BS to claim that every hocus pocus game with a thermal engine or a neural network is important for the progress of science.

The real advances are way more localized when it comes to the identity of the humans who do it (the credit); when it comes to the location where the advance takes place; and when it comes to the timing (often seconds although some calculations or experiments that take a much longer time are sometimes important, too). This point – the localization of the advances and their abrupt character – is something that Dijkgraaf seems to deny. He envisions a brave new world of pseudoscience where everything is diluted over people, space, and time. All people are equal, all moments and places are equal, and everyone, every place, and every moment must be constantly praised. They are not equal and even most alchemists would agree with me that they are not. In this sense, the alchemy-like neural network hocus pocus is being mixed with the radical left-wing delusions such as "equality" and their plan is to cancel the respect for real achievements of mankind and science, and to reverse science to an epoch of muddy pseudoscience and deceitful idiotic promises (psychopaths talking about the zero carbon emissions are another category of these dishonest pseudoscientists who place promises convenient for their wallet above the scientific truth).

A person playing with a computer is generally not a scientist; the neural network mumbo jumbo is not science; random processes or games are generally not scientific progress even if they are generously funded; real advances are localized and often underappreciated. Science has to be honest, impartial, non-corrupt, avoid promising things that seem impossible, and it has to find some laws that must be understandable by humans.

No comments:

Post a Comment