"We Will Not Notice How The World Will Be Taken Over By Artificial Intelligence" - Alternative View

"We Will Not Notice How The World Will Be Taken Over By Artificial Intelligence" - Alternative View
"We Will Not Notice How The World Will Be Taken Over By Artificial Intelligence" - Alternative View

Video: "We Will Not Notice How The World Will Be Taken Over By Artificial Intelligence" - Alternative View

Video:
Video: Not If, But How Artificial Intelligence Might Take Over the World | Hugh Baillie | TEDxISKL 2024, May
Anonim

British company Cambridge Analytica helped Donald Trump win the US presidential election using Big Data technology and personalized online advertising. Prior to that, the same firm worked with Brexit supporters in Britain, and has now signed a contract with the French National Front. Is it true that unexpected voting outcomes in different countries are not a failure of sociology, but a victory for a new type of sociology? The interlocutor of Radio Liberty Michal Kosinski, whose research is indirectly related to the activities of Cambridge Analytica, believes that this is an exaggeration, but Big Data technologies and a decrease in privacy promise the world global changes.

Back in late September, in the midst of elections, speaking in New York City at the Concordia Summit, an annual event for influential politicians and businessmen from around the world, Cambridge Analytica founder Alexander Nix talked about how the new technology has improved the effectiveness of the campaign of Trump's Republican nomination, Ted Cruz., by the beginning of the race, almost unknown candidate. Nix talked about how, for example, people of different temperaments should present a candidate's opinion on the law on the free distribution of weapons in different ways: fearful people with a high level of neuroticism can think of weapons as a source of safety, and rich conservative extroverts can be shown a picture of a duck hunt.

Later, Cambridge Analytica helped Trump himself, and, according to Nyx, for a relatively modest reward, in total, about $ 15 million. By analyzing the data that all Internet users leave on the network, the company's specialists managed to find non-obvious connections and patterns (for example, that people who prefer American-made cars are Trump voters) and use them to target advertising on the Internet, to personalize messages. outgoing from the headquarters of the Republican candidate. Residents of the Little Haiti quarter in Miami were shown information about Hillary Clinton's refusal to participate in the relief of the earthquake in Haiti, African Americans - a video where Clinton compares black men to predators. Trump volunteers received a detailed profile of the residents of the houses,where they were going to conduct direct campaigning.

It is difficult to say how much of an impact Cambridge Analytica's work has had on the outcome of the US election, but this campaign is not the only example of the winners being those who hired the firm. For example, back in 2015, the radical supporter of Brexit Nigel Farage began using Cambridge Analytica. The Swiss edition Das Magasin, which recently published an investigation into the activities of Cambridge Analytica (you can read it in Russian here), says that the company has now received a contract with the French National Front, in addition, some political forces in Switzerland, Germany and perhaps Italy.

One of the heroes of the Das Magasin publication is Michal Kosinski, a researcher of Polish origin, formerly Deputy Director of the Center for Psychometrics at the University of Cambridge, and currently an assistant professor at Stanford University in the USA. For several years, Kosinski and his Cambridge colleagues have been developing a system that, based on the user's activity on a social network, compiles a detailed psychological profile of a person. The system is capable of not only describing character traits, but also predicting, among other things, gender, sexual orientation, skin color, and even political preferences of the user.

Swiss journalists say that in early 2014, Kosinski was approached by a colleague, Alexander Kogan, who invited the young researchers to apply the model he had created to analyze several million profiles of American citizens at the request of a certain company. As it became clear later, this company was associated with Cambridge Analytica. Kosinski declined to cooperate, but the methods Cambridge Analytica used during the referendum in Britain and the US presidential elections are at least very similar to the models invented by Michal Kosinski.

Radio Liberty phoned Kosinski in California and asked him what exactly can be said about a person on his "digital trail", how much to trust the loud statements of Cambridge Analytica, whether the war for privacy is lost and whether artificial intelligence is more dangerous than Donald Trump.

- What is psychometry? What do you do as a researcher?

Promotional video:

“Psychometry itself is an ancient science, I think it is two or three millennia old. In essence, this is the science of psychological measurements, attempts to establish as accurately as possible various psychological aspects, personality, intellectual abilities, happiness, tendency to depression, and so on. Traditionally, such measurements were made using questionnaires, psychological tests, but relatively recently I and some other psychometrists realized that they can be done by assessing a person's digital footprint - and then no questions can be asked, no questionnaires and tests are needed. And this is a revolutionary moment. The importance of the Cambridge Analytica story, which Das Magasin wrote about, is not really that the company helped Trump in the election race. This is a commercial company, they have technology, they want to make money, everything is clear here. What matters isthat if earlier you wanted to compose someone's psychological profile, you had to ask the person to fill out a questionnaire, take a test - and the subject understood well that right now, at this very moment, someone is measuring his psychological characteristics. And now you can do the same, but the person does not recognize that his most intimate features are being evaluated and measured right now by someone. Just look at the digital footprint: social media posts, likes, web browsing history, search history. Based on this data, an incredibly accurate psychological portrait can be compiled. On the one hand, it seems like something scary, on the other, it can be very useful. For example, some Internet platform may suggest the most suitable job for your character and abilities, or recommend a film,which you will surely like. This is normal. But when you open your Facebook page and see ads there that are personally targeted, targeted to you based on your detailed psychological profile, which was compiled without your knowledge and consent, this is no longer very normal.

- So, personally, could you make my complete psychological profile, knowing only my name, based on the footprint that I left on the Internet?

“I’m a scientist and I just wouldn’t do it. If you volunteered to help my research and provide your data, I would gladly include it in my database, anonymize it and promise never to sell or share it with anyone. But many companies, of course, constantly collect information about you without asking and use it, for example, for targeted advertising.

- This information, of course, is not stored in one place - one company knows my geolocation data, another - financial transactions, the third - likes on Facebook. Can all this be put together?

- Well, first of all, this is usually not necessary. Usually, one thing is enough to build an accurate profile - your interests on Facebook, for example, or your browsing history from your browser. Secondly, companies that collect information about you usually exchange it with each other or simply sell it in the general data market. There are projects like Acxiom that put everything together and sell it to firms like Cambridge Analytica. You can just come and say - I need data on 10 million Americans from such and such a region, and they will gladly sell this data. Thirdly, in order to run a campaign with personal targeting, it is not at all necessary to have individual user data. You can use the so-called behavioral targeting. For example, you cannot ask Facebook to show certain ads to all people,prone to a certain type of behavior. But you may have a model linking this type of behavior with some kind of psychological profile, and you could even build this model quite honestly - you paid several tens of thousands of people a small fee to fill out the questionnaire. And when the model is built, you ask it: how do I find emotional extroverts? The model replies - no problem, these are those who liked such and such ten books, such and such ten films and such and such ten musicians. Now you go to Facebook again with your ad, only you no longer need to ask to show it to emotional extroverts, which Facebook would not do. Instead, you're asking to show ads to those who have a specific set of likes. As a result, it turns out that you did not have any personal data,but you ran personally targeted advertising.

- Are there companies that seek to compile a psychological profile of all of humanity, at least everyone who leaves digital traces?

- Hardly anyone thinks in such global categories, but to some extent Facebook, Google, Microsoft, Visa, Mastercard, the same Acxiom are engaged in this. And all this data is widely traded in the market.

- That is, somewhere on the market, for example, my psychological profile is also sold, and it is incredibly accurate. You said in a comment for Das Magazin that just a few likes on Facebook are enough for the system to recognize you better than your closest friend. Is it really so?

- Yes, there was an interesting study about this. Ten likes (interests) are enough for the system to better recognize your personality than a work colleague, and by 230-240 likes the computer will know more about you than your spouse.

“But what exactly does it mean to know more?

“This means that if you ask the computer to fill out a psychological questionnaire for you, it will make less mistakes than your wife.

- These are the answers to questions like "Are you afraid of the dark"?

- Yes, typical questions of psychological questionnaires. So really, somewhere on the market you can buy very accurate information about your personality, including your political bias, your religiosity, your sexual orientation, your IQ. You can roughly understand what is known about you based on your interests on Facebook using the Applymagicsauce.com site, but this, of course, is far from the complete picture.

- Let's turn to the story with Cambridge Analytica. Tell me, you really have nothing to do with this company?

- No, and never had. I heard about its existence from the press.

- Alexander Kogan, who, according to Das Magazin, offered Cambridge Analytica psychological profiling technology, was your colleague?

- Yes, he was my colleague, or rather, he was an assistant professor at the psychology department of Cambridge University at the time when I was a graduate student there. But our paths diverged a long time ago, he founded a small company and, as far as I understood from the press, sold Cambridge Analytica data, and I stayed in the academic world.

- Can we assume that Cambridge Analytica is using the models you have developed?

- I would say that they use similar models, but, you see, to develop such a model, you do not need particularly deep science, there is no magic in it. Anyone with basic programming skills, some money, and a computer connected to the internet can do this - in fact, that's what makes the problem so big. The most standard statistical methods are used here, nothing special. The main idea of my publications is that it is easy, it is only important to change the focus: questionnaires and tests are no longer needed for psychometrics, it is enough to have a digital footprint of a person. And this gives huge benefits, you can improve marketing, career planning, methods of psychological assistance, and much more. But the same technology can be used against humans. I am sure that Alexander Kogan was aware of my research and,although I can only judge this from the reports of journalists, he did exactly what I suggested and sold the result to Cambridge Analytica. I have said many times and clearly indicated in my articles that I did not want to inspire anyone to do this kind of activity, moreover, I am sure that people were engaged in digital psychological profiling long before me, they simply did not talk about it as actively as Cambridge Analytica. I just call for the development of policies and procedures that would determine the ethical use of this technology.that people had been doing digital psychological profiling long before me, they just didn't talk about it as actively as Cambridge Analytica. I just call for the development of policies and procedures that would determine the ethical use of this technology.that people had been doing digital psychological profiling long before me, they just didn't talk about it as actively as Cambridge Analytica. I just call for the development of policies and procedures that would determine the ethical use of this technology.

- Do you think Cambridge Analytica really could significantly influence the results of the American presidential election?

- Honestly, I do not know. It is clear that Cambridge Analytica is interested in trumpeting its success as loudly as possible, but in the end the election results are determined not by Big Data methods, but by candidates, and even, as it is now especially clearly seen, voters who do not go to the polls. Perhaps the activity of Cambridge Analytica was the very last straw that predetermined the final result, but most likely, Alexander Nix himself does not know for sure. On the other hand, we must be aware that there is technology, and it is almost impossible to defend against it. You can stop using Facebook, but you will still be writing emails. Decide to use pigeon mail - you can't do without a credit card. Maybe countries could pass laws restricting such practices,but that may not work either. By the way, I don’t know for sure, but I heard that what Cambridge Analytica did for Trump was contrary to current American law, and therefore they did direct analysis of the data in Britain. You know, I am a scientist, I do psychometrics, I am not an expert in politics, democracy and freedom, but it seems to me that we will have to accept the fact that there will be no privacy left. Instead of getting involved in another battle for privacy, it is worth recognizing that the war has already been lost, and it is better to attend to the fact that the world becomes a favorable environment for a person devoid of privacy. You know, I am a scientist, I do psychometrics, I am not an expert in politics, democracy and freedom, but it seems to me that we will have to accept the fact that there will be no privacy left. Instead of getting involved in another battle for privacy, it is worth recognizing that the war has already been lost, and it is better to attend to the fact that the world becomes a favorable environment for a person devoid of privacy. You know, I am a scientist, I do psychometrics, I am not an expert in politics, democracy and freedom, but it seems to me that we will have to accept the fact that there will be no privacy left. Instead of getting involved in another battle for privacy, it is worth recognizing that the war has already been lost, and it is better to attend to the fact that the world becomes a favorable environment for a person devoid of privacy.

- It is curious to what extent building a psychological profile, using personal targeting can affect the real world.

- From a scientific point of view, it is difficult to argue about this, because it is difficult to set up experiments. That is, if you sell some kind of shower gel, you can compare the number of clicks, but it is very difficult to evaluate, for example, the effectiveness of Big Data methods in a political campaign, which is more like a brand promotion than a sales campaign. We did some experiments, not with politicians, because I think it is unethical, but with specific consumer products, and it turned out that personal targeting, that is, using an individual advertising message based on a psychological profile, doubles the effectiveness of advertising. And this is an incredible result, usually the struggle is to improve efficiency by some three percent, and even this saves companies hundreds of millions of dollars. And then twice!

- It's strange that you haven't become a billionaire yet.

- I already have a wonderful life. By the way, there are many psychological works that prove that happiness does not require so much money, well, you need, of course, enough for a vacation. In fact, everything is probably not as simple as it seems to me as a scientist, I'm just ready to talk about these things openly, while others are working on real projects without telling anyone anything.

- To me, the idea of psychological profiling based on a digital footprint seems quite natural, I'm sure a lot of people already use it in advertising. Our world may have already changed a lot, we just don't really notice it yet.

- Probably, and I rather welcome these changes. We had inventions and more terrible than Big Data, we coped with nuclear energy - we can somehow cope with this, although, probably, not without sacrifices. As I said, it seems to me that the war for privacy has already been lost, we live in a post-privacy world, which means that it is worth investing in ensuring that everyone has equal rights and opportunities, so that no one wants to hide anything. Hopefully, more tolerance can solve the problem of lack of privacy.

- But the problem is not only that others will learn all the secrets of a person and begin to treat him differently, but also that a powerful tool for manipulation has appeared.

- With this, in my opinion, education should cope. It is easy to manipulate those who know little, who read little, who travel little. Such a person can be told that there is no war in Syria, and indeed all Syrians are bloody criminals, and he will readily believe in it.

- In my opinion, educated people can also be manipulated, with the help of vanity and sex, for example.

“I don’t think so. So you probably know your wife quite well, does this make it easy for you to manipulate her?

- I don’t know, I probably simply don’t have such a goal, but there will certainly be many people who would be interested in manipulating me as efficiently as possible so that I go and vote, for example, for Putin.

- Of course, there will be, and this happens all the time, all advertising, all marketing is manipulation of this kind, and I don't see what we can do about it. Yet it is difficult to manipulate educated, open-minded and happy people. And it is too late to give up Facebook, e-mail, credit cards, and it is simply stupid. Facebook is a wonderful thing, so I've talked a lot in my life about the risks of using it, but I use it myself, and with pleasure, and I would even pay for it if I had to.

- By the way, you were offered a job on Facebook?

- Yes it's true. I'm fine with my job at university, but actually, for a behavioral researcher who wants to change people's lives for the better, a career at a company like Facebook or Google is a great way to go. I come up with models and study them, and if I worked on Facebook, I could actually use them to change the lives of millions of people, and it is not at all necessary to sell them something.

- It's scary that you can change the lives of millions of people with a couple of clicks.

- Indeed, in order to influence millions of people, you used to be an authoritarian dictator, but now you can be an engineer working for some Internet company. But this is not scary. All of these algorithms - Google searches, recommendation systems, and so on - are based on the most basic math, but it applies to huge amounts of data. And in the end, no one understands how it all essentially works, not even the developers themselves. Here in Russia there are some laws that everyone considers bad, but these are obvious laws that can be corrected. And computer algorithms self-learn and become so complex that if they work somehow wrong, it is not clear what and where to fix them.

- Are we so far from the moment when some such out of control, inaccessible to our understanding system will arrange a nuclear war?

- You know, so far Donald Trump has much more opportunities and chances to destroy humanity than Big Data, but in the future … Look, there are pigeons and ants, do you think they know that the world is ruled by man? Are they guessing that Donald Trump can press the red button? Of course not. It is worth remembering this, because when artificial intelligence begins to rule the world, we may well not notice it in the same way. After all, we can no longer comprehend and analyze the volumes of data that we ourselves produce - and we have to trust the computer. You may ask the machine: can you predict something based on this data? She will answer: yes, of course. You ask: how do you do it? She will say: I'm sorry, but you won't understand. And all this is already a reality. You know that American courts use computer systems,deciding on parole for prisoners? Computers decide whether to give you credit or not. The computer lands a plane, which is great, because people can easily cause trouble. The lives of 800 Boeing 787 passengers are in the hands of a computer, not a person, and this is today's reality.

Recommended: