Artificial intelligence guarding Spanish order
Recently, several articles have been published stating that artificial intelligence ultimately did not change anything. Of course, this is not true. The development of AI leads to an increase in the specific productivity of each individual employee, and little by little all companies are optimizing and reducing staff.
I started looking for cases where AI is actually used and it has an effect.
Perhaps the most interesting case is the story of how the Spanish police are actively introducing AI into their activities. For example, back in 2021, VeriPol was introduced, which verifies reporting.
In Spain, police officers often file false reports to open cases. Usually these are cases of moderate severity. That is, the policeman says that there is such a case, investigates it and closes it, receiving encouragement.
During the pilot project, the model accuracy was 83%. The accuracy was expected to be 91%. But the system was accepted and used. Perhaps after 3 years it became more accurate. However, even at the start, the AI turned out to be 9% more accurate than the invited police expert.
The police did not stop there and introduced the VioGén algorithm, which assesses the likelihood of committing a crime. The algorithm has been used since 2007, that is, it was a rather bold step.
The algorithm began to be used to analyze situations related to domestic violence. It looks like this. A Spanish policeman arrives to respond to a call where some kind of family squabble is taking place. And here the Spanish police are no different from the Russian or any other. They can only make a suggestion, and will investigate and punish only after the crime has been committed. I think everyone remembers the phrase that made the headlines in the media: “When they kill you, then call.”
The Spaniards decided to try to act preventively. The VioGén algorithm contains 35 questions, based on which it generates the probability of risk. If it shows that the risk is low, then the police leave, and if it is high, then they speak sternly to the offender and only then leave.
The world’s media loves to kick VioGén for its mistakes. Thus, The New York Times cites several stories where the algorithm said that the risk was low, and then the woman was killed or beaten to death. However, here I would like to cite some positive statistics published by the local ministry. During its operation, the algorithm studied 814,000 cases of domestic violence and was only mistaken in 0.03%, that is, 245 women were killed.
And here, too, we can talk about whether it is worth introducing artificial intelligence into crime investigation, because, firstly, it seems to remove responsibility. The police say that they are not guilty, but the algorithm said that the risk is low.
Secondly, skills seem to be dulled and a formal approach prevails: the police come, ask questions, enter them into the tablet, look at the result and leave.
Thirdly, here you can insert a dystopian part, when a person has not yet done anything – well, he yelled, well, he kicked a couple of times (tell us how domestic violence happens in your country) – but he was immediately tied up and taken to the monkey house. It’s scary to imagine what will happen if such a system is extended to other areas of life.
How do you look for information?
I remember being taught how to use catalogs in the library. It felt like some kind of hack into the entire library system. For those who have forgotten how everything works, let me remind you that, in addition to the alphabetical catalog, there are also subject, systematic, geographical and regional studies. You come like this to the library quickly slap-slap-slap, and at school you have the best essay with the largest number of sources.
Then I got the Internet and search engines. Queries could be written just like that, but a much better effect was obtained when using syntax. Now, by the way, I use search operators minimally, because Google has learned to understand exactly what the user needs. The only thing left in my arsenal is quotation marks for precise wording, the exclusion of specific words, and recently I used the “loc” operator to look up the cost of mobile communication tariffs in Venezuela.
But you shouldn’t forget Yandex operators. It seems that Yandex understands me worse than Google, but at the same time, Yandex is necessary for searching the RuNet.
However, I would like to draw your attention to the fact that in the last year there has been a transition towards neural networks. And, to be honest, it happened somehow quickly. Unexpectedly, it turned out that my most technically backward friends (I called them that; they themselves may have a different opinion) are making full use of the achievements of neural networks: they generate pictures, create music, and use them to search for information. And they often do this in ways that I would not even think of. So, one acquaintance on a tourist trip simply dictated to ChatGPT the sights he wanted to visit and asked him to come up with a program. And the neural network did the job: it plotted the route, indicated that it was better to go to this museum on Tuesday morning, because it closes at 2 p.m. and is closed on Monday, and found a place for lunch. If you connect a bank card here, you’ll have a full-fledged personal assistant who will come up with a route, order tickets, and organize logistics.
Another amazing example for me is how often the Summary function is used (this is when a neural network reads a document for you and gives a short summary). How can you entrust something like this to a neural network? What if she reads it poorly and misses some important point?
On the one hand, it is quite clear why search through neural networks is on the rise. Because of search engine optimization and paid promotion, searching for information is relatively difficult because you have to break through the promoted pages. Often the necessary content is on the second or even third page of search results. Although, according to various studies, from 75% to 94% of users do not look beyond the first page. However, most of these studies are carried out by marketing firms involved in optimization, so the results may be biased. Tell us in the comments how you search for information.
I have written more than once that checking the success of an invention is quite simple: you just need to ask the question whether the positive opportunities from the invention outweigh the disadvantages that need to be dealt with. For example, there are a lot of hassles with a car (from maintenance to pollution of the planet), but there are many more positive opportunities from its use, and this is appreciated by people all over the world.
In this case, the neural network quickly generates a complete answer for you. And users are willing to accept that the answer may be incomplete, inaccurate, and sometimes erroneous. But you don’t have to waste time on a tedious search.
It is likely that we are moving in the direction where the search bar will become a professional tool for a certain circle: journalists, analysts, sociologists, etc.
For example, I love search engines (Google, Yandex, DuckDuckGo, Presearch), as this is an opportunity not only to answer a question, but also broaden my horizons along the way. It often happens that by asking one question, you find a lot of new and useful information.
However, often people are looking for answers to specific questions and do not want to think about how to formulate a query and which search engine will give the best answer. It is likely that this will lead to some transformation of the Internet, as many sites will lose traffic. For example, our website, because the user, instead of independently reading reviews of several smartphones, will ask the neural network about this and make a decision based on its answer. In turn, the site will receive a pretty penny for the fact that the answer is based on its data.
By the way, The Atlantic magazine published a funny article in which the author sighs that AI is killing curiosity. Like, a person asks the AI a question, gets an answer, and that’s it. And I immediately remembered how exactly the same thing was said about the Internet. This was the time when Yandex, armed with the slogan created for it by Artemy Lebedev, “Everything can be found,” powerfully promoted itself. I think it was 2000. And I listened to a program on the radio at breakfast, as a guest expert argued that with the Internet, people will know less. And he gave, in particular, an example about learning a foreign language. Without the Internet, you take a dictionary, leaf through it, come across other interesting words while looking for the translation you need, then try to translate the sentence yourself. And Yandex immediately gives you a translation and thus deprives you of the happiness of knowledge.
But it seems to me that with the Internet, people, on the contrary, have become a little wiser, because knowledge is “at their fingertips” and curiosity can be immediately satisfied, and not everyone has an encyclopedic dictionary at home. Yes, and too lazy to rummage through it.
Conclusion
Perhaps the main danger of AI is that people want to shift all responsibilities to it, so that someone else can make decisions for them. It’s not for nothing that articles and reels on social networks are so popular, where people tell how they entrust AI to make all decisions for a certain amount of time, and then they follow them and, as a conclusion, tell that they have become more productive, get better sleep, etc.
But in fact, AI is nothing more than an assistant that should help a person do more, but not make decisions for him.
Source: mobile-review.com
