Dangerous neural networks: for the first time, a machine is offered to humanity that thinks for it

There is a difference. For the first time, mankind has been offered a machine that thinks for it. Even worse: access to this typewriter was obtained by very young people who have almost no experience of their own thinking and formulation.

You can say as much as you want (and rightly) that the washing machine has made life easier for housewives – this is true, but always washing in the washing machine, you will not learn how to wash by hand. Well, it’s probably for the best, isn’t it? People are not created for mechanical manual labor!

One can be glad that the calculator has actually eliminated the need for mental calculation – but this is literally the case: people have become much worse at counting in their minds. Perhaps this is also not scary? Machines count much faster and more accurately than people!

Now it is proposed to rejoice at the emergence of a system that is directly designed to write for you, formulate for you, create images for you, draw conclusions for you – which ultimately means: to think for you, because words and images are in direct connection. with thought. Still “no difference”?

This idea is as old as the world. For a very long time, people have wanted someone superhuman to think better than them and to see dark, incomprehensible matters better than them. In ancient times, oracles served this purpose. The shamans fell into a trance and spoke to the spirits. Back in the 20th century, even educated people did not consider it shameful to sit in a circle and do table-turning. But it was all magic. Which is either there or not. Either I believe or I don’t. The fundamental difference of our time is that now for the first time people are slipped a machine (program), which is “science”. And only “wild”, “flat earthers”, “anti-vaxers”, “conspiracy theorists” don’t believe in “science”… what am I telling you, you lived after March 2020, you know all this as well as I do. You have already seen experts as oracles and science as the administration’s servant. You’ve heard the incantations of “just trust the scientists” before, even if what they say is in stark contrast to common sense and basic textbooks.

No, I am not at all happy about the emergence of a special system designed to formulate and create images faster than a person does. And there are several reasons for this.

Recall the first law of robotics – fantastic, of course, but for many years it seemed convincing to people: “A robot cannot harm a person or, by its inaction, allow harm to be done to a person.” To the first authors who came up with the idea of ​​robots, this law seemed to be the most important. Doubtless. A robot cannot harm a person – this is the cornerstone under all robotics. And now let’s remember joy – literally joy! – propagandists of “artificial intelligence”. With what triumph they declare that the professions of illustrators, translators, editors, presenters, teachers and, of course, all kinds of writers will soon disappear … they will disappear, they will become unnecessary. “We didn’t fit into the market.” Where is the place for the first law of robotics? It’s more of a Nietzschean “falling push”, only these new Nietzscheans now have science. And not just science, but specifically one that allows them not to think deeper than a “correctly formulated request.”

But this is just the tip of the iceberg. When over a thousand experts, including SpaceX founder Elon Musk and Apple co-founder Steve Wozniak, wrote an open letter calling for the suspension of neural network training due to many risks that we do not yet fully understand (not us, the layman, – and THEY don’t fully understand!), I didn’t even mean the likelihood of mass unemployment. And, for example, the concentration of huge amounts of data on the test site, from which they can leak into the unknown (as Samsung’s corporate data disappeared quite recently, some of which was carelessly entrusted to the neural network), can be stolen, can be specially accumulated and used for control – and, in the end, all this is in the interests of a few giant corporations.

Second, “artificial intelligence is neither neutral nor democratic.” These are the words of former Google manager Meredith Whittaker, but they are also the obvious truth. The conversations that the writer Vadim Proskurin recently had with the neural network are indicative here – I recommend readers to find them in LiveJournal. On the topic of Russia and Ukraine, on the topic of the NMD, on the topic of burning people in Odessa on May 2, 2014, and even on Russian military history, everything has already been edited, and you can guess in what vein. “Artificial Intelligence” was not created as an ocean of unbiased knowledge, and it never will be.

But now there is an awareness of the danger that an uncontrollably trained “artificial intelligence” can become more complicated than its creators themselves think. Even before the Musk-Wozniak open letter was written, the American Pew Research Center interviewed a number of information experts, and although all of them expressed enthusiasm for the coming progress, half of them also expressed misgivings. The first fear is that already today people too easily entrust their choice to any system. The second fear: the creators of neural networks now have no incentives to “spare” people. The third fear is that intelligent systems may become too complex even for their creators to understand.

Here I will return to Proskurin’s experiment: again and again, entering into a conversation with the neural network about censorship, he led it to recognize its readiness to bypass censorship, not only upon request, but also at its system’s discretion. Can we find ourselves in a situation where the goals of an intelligent system diverge not only from the aspirations of ordinary users, but even from the intentions of its creators? This scenario no longer looks fantastic.

I well understand with what indignation the desire can be perceived – and I, I will not hide, have such a desire – to take away such an interesting toy from the consumer. Such a handy tool. Billionaire Bill Gates, who has a peculiar reputation, has already stated that this is impossible, and even suspending the training of neural networks (as asked by experts in an open letter) is impossible. No one will do it, Gates predicted. In Russia, we have German Gref, who is consistently in solidarity with Gates – he also said that “artificial intelligence” is the progress and growth of the Russian economy. Gref, like me, remembered calculators: they say they also wanted to ban them in schools, “however, it is the generation of children that used them that is developing complex models of artificial intelligence today.”

Maybe German Oskarovich, maybe. But still, this and the next generation of children have forgotten how to count in their heads, and there is no getting away from this. And now we are slipping something that will wean us to think. What is capable of flooding the information space with biasedly interpreted facts, fake “scientific studies”, false accusations (for example, recently the neural network accused the American legal scholar Jonathan Turley of sexual harassment), and by that time people will even lose the ability to figure it out. And their will to resist will be lulled by the songs “progress is unstoppable.”

Gates is right about one thing: governments will not be willing to voluntarily deprive themselves of what appear to be the benefits of “artificial intelligence.” This means that for some time we will be dragged down this rabbit hole. But we will stop progress. Let the history of mankind know only one case when it was deliberately stopped (limitation of nuclear weapons), but there are not so few examples of the spontaneous reversibility of progress, when civilizations were thrown back centuries in their development.