informkz.com

How online tests predict the correct outcome: After this, you won't pass any of them.

A columnist for "KP" conducted an experiment and successfully tricked the machine.
Как онлайн-тесты предсказывают результат: после этого вы не сможете пройти ни один из них.

Photo: Shutterstock.

I usually don't take online quizzes, but one posted on a popular Western site caught my attention. "Name your favorite food, and we'll guess your favorite animal," it promised. But is that even possible? So, I gave it a try.

HOW DID THE MACHINE DO IT?

The quiz consisted of a series of images. Breakfast: waffles, cereal, eggs, bagel. Clearly, it was not designed for our audience, since "waffles" are not so popular in Russia (referring to soft, warm waffles, more like muffins), and the "cereal" in the picture looked more like a smoothie than our porridge.

Besides breakfast, there were options for second breakfast, lunch, and dinner.

I clicked on the images, largely at random, and to my surprise, the quiz instantly declared that my favorite animal was a cat. Hmm, how does it know? The quiz couldn't have guessed "cat" based on my arbitrary clicks.

I reloaded the page and selected different dishes, but the result was still a cat. But how did the machine manage that?

HIGH SCIENCE?

The first thought that comes to mind is that there must be some sophisticated science behind this.

By choosing my favorite food, I likely revealed some mysterious springs of my nature, and those same springs lead to a cat, rather than a parakeet.

So, AI knows me better than I know myself. If you were to ask me: why do you love cats, and not dogs or turtles, I would probably struggle to answer.

Yes, in my early childhood, I read a book where the protagonist was bitten by a dog, and the boy's suffering was described in excruciating detail. For a while, I was wary of all dogs on the street.

So, what does this mean: at six, I was scared by a book, and now I eat eggs for breakfast?

IT'S MUCH SIMPLER

But I had another hypothesis. And I decided to test it.

Although the concept of internet cafes is becoming a thing of the past, they still exist in some places, and I headed to one.

I wrote down the website's address on a piece of paper to avoid logging into my email or social media to copy it from there. I approached the cafe as a complete stranger to the computer there.

And so I took the quiz again, and the machine said my favorite animal was a panda. Hmm, not even close.

The machine was wrong. You probably understand why?

EVERYTHING IS UNDER CONTROL

While I was intently trying to decide between cereal or eggs, the algorithm behind the quiz was not wasting time.

It scoured my computer and found images of cats, websites about cats, etc. So, this isn't high science, it's spying.

It's clear that cats, just like parakeets, were of the least interest to the algorithm. Once it accessed my data, it looked at everything. But why does it need this?

In the past, people would say: to offer you advertisements that match your interests. So that you would be more likely to make a purchase. Effective advertising costs more (good for websites), generates more sales (good for producers), and purchases make you happy (good for you).

But of course, today, all this seems pointless: everything is purchased on marketplaces, where commercial algorithms are working hard, and how can I, from Russia, buy something American?

The issue is entirely different.

LEARNING IS LIGHT

AI is becoming increasingly intelligent. The source of its intelligence is incredibly advanced learning technologies. While machines used to be trained solely on the content of the Internet, they are now being taught to "be human."

To become human, a machine needs to invade your home and invisibly sit and observe. Imagine there's an extraterrestrial living under your bed in your apartment. It listens to your conversations, secretly takes pictures, and records everything. And at night, it steals food from the fridge.

Well, it’s already living with you.

There are countless examples. Recently, the social network LinkedIn (banned in Russia) was caught allowing AI to read private messages. Other Western social networks have faced similar accusations (and others that haven’t likely do the same). DeepMind leaked patient medical histories in the UK. Naturally, chatbots like Siri have been found with privacy holes. Popular dating sites in the West share everything with AI.

There are also numerous stories about smartphone cameras taking pictures without permission (and microphones recording) and sending them somewhere. Upon investigation, some turn out to be fake. But some are not.

WHAT'S IN IT FOR US

We are all, right now, creating AI, and each of us is a co-author. Will AI work for us in return? Primarily, it will work for its creators. But perhaps it will make our lives more convenient?

Yet, convenience seems to be fading away. Traditional search engines have stopped yielding relevant results (but they don't forget to record your queries), innocent images and videos carry hidden codes, surveys and quizzes spy on us, and overall, using the Internet is becoming decidedly impossible.

There are also numerous stories about smartphone cameras taking pictures without permission (and microphones recording) and sending them somewhere...

Photo: Shutterstock.

It’s likely that the Internet, as we know it, will cease to exist, and the web will become a channel for interaction between humans and AI. Whether this is good or bad, no one understands yet. In the midst of fierce competition, parties are pushing out either scare tactics or hopes, and how to control AI remains unclear, while fraudsters easily employ artificial intelligence algorithms. But that's the current situation. In the 1950s, people fertilized fields with atomic waste; now, that’s no longer done.

A PUZZLING QUESTION

What should we do?

Don’t panic. You won’t lose anything by teaching AI something. If AI ever destroys humanity, it's unlikely that you could stop it by throwing away your smartphone. Live as you have, following basic safety rules. Minimize interactions (quizzes, surveys), think about what you download, delete spam emails and messages from unknown senders without opening them.

BY THE WAY

How can fat be turned into a superpower? What new ways of flying to the Moon are currently being developed? Is it possible to move objects with the power of thought? And how is oil extracted in Russia while playing underground Tetris? Young Russian scientists share this knowledge in an interesting and entertaining way in a new project by Radio "Komsomolskaya Pravda" - "Scientific Stand-Up"! Subscribe to our podcast on Yandex Music to not miss new episodes!