Close your eyes and try to imagine what artificial intelligence looks like. What do you think? A terminator robot with ominous red eyes, a blue human brain floating amidst luminous numbers and formulas, or a white humanoid robot mimicking Rodin’s “The Thinker”? These images are the most stereotypical images of AI on the internet.

Try typing the phrase “artificial intelligence” into Google’s image search engine and you’ll see for yourself.

The number of posts about AI on the Internet is growing, and with it myths about it are being consolidated, which owe their appearance largely to the appearance of information about AI in the mass media – clickable headlines and image selection. A few years ago, one of the artificial intelligence researchers Daniel Leifer he even created a site dedicated to the myths about artificial intelligence that he discovered by analyzing media reports for more than six months. These myths range from extreme optimism that AI is a magical miracle that will save the world, to extreme pessimism that AI is a weapon of mass destruction.

We will talk about our 3 most important myths spread in the media about artificial intelligence.

MYTH #1: AI HAS FREEDOM OF MOVEMENT

“artificial intelligence opened a new antibiotic”, “AI Wrote Text based on biblical motifs”, “AI exceeded doctors predict heart attacks» – mass media often write about artificial intelligence as the main protagonist of scientific discoveries or achievements.

Leifer highlights the fallacy of such a formulation, as it creates the illusion that artificial intelligence is capable of performing the same actions as humans. In this way, abilities are attributed to artificial intelligence that it lacks, especially the ability to act independently or freedom of will. When people see the “AI” banner opened a new antibiotic,” they may have a picture in their brains of how artificial intelligence independently conducts scientific experiments and analyzes data to discover new drugs.

But such an idea is far from reality and hides the role of man in achievements. The first are people who develop some sort of artificial intelligence system and program and calibrate it to achieve specific results. The latter is usually the result of some kind of discovery or achievement, analyzing a large set of data with the help of artificial intelligence. In order to do this analysis, people had to first gather the data, make sure it was adequate, and make sure it was a balanced representation of the problem to be solved. All these processes take years, if not months, of painstaking human work. Finally, analysis results published by AI need to be interpreted and validated correctly because the analysis may be inaccurate.

That’s why researcher Leifer emphasizes that AI is nothing more than a tool we use to achieve our goals, like a pen, calculator or computer. We usually prefer to say “I wrote/wrote this material with a pen” rather than “this pen wrote the article”. Leifer wants us to keep that in mind when we see another sensational headline that AI has accomplished something.

MYTH 2: AI IS LIKE HUMANS

In addition to headlines, an important element in the presentation of news in mass media is visual. digital sociology researcher Lisa Talia Moretti and founder of arts and innovation consultancy “Culture A” Anne Troyer Rogers He analyzed more than 1,000 news stories and found that artificial intelligence was mostly portrayed as humanoid.

The researchers note that AI is equivalent to white or silver robots with the same eyes, mouth, torso, and hands as humans. Sometimes, they show signs of aggression such as bare teeth and glowing red eyes. However, AI is not always closely associated with robots. For example, ChatGPT was built on the basis of artificial intelligence, but it is a chatbot, so it would be wrong to depict it using a cyborg image. Even when a robot, such as an object recognition robot, is powered by artificial intelligence, it often looks nothing like a human. Additionally, some images show technologies that do not currently exist and are unlikely to be created in the future, such as a humanoid robot listening to a patient with a stethoscope.

According to the researchers, another wrong aspect of comparing artificial intelligence to humans is in pictorial actions. Often, robots are portrayed as if they are deep in thought or solving complex math problems on a blackboard. In this way, the illusion is created that AI is used for extremely complex tasks rather than humans, thus emphasizing its superiority over humans. This can fuel unrealistic fears and expectations about AI advances – either human-level intelligence or even “superintelligence” that could wipe out humans in the near future. But the main goal of artificial intelligence developers is not to replace a person, but to expand his abilities. For example, creating text that can be presented during high-level political speeches or processing large volumes of text to analyze customer feedback.

The “Better Images of Artificial Intelligence” research group also believes such false illustrations, as well as false headlines, obscure human involvement and reputation in the achievements and discoveries made with AI. They have developed a gallery of more relevant AI images showing the realistic state of development of this technology. Their aim was to emphasize that AI is a technology made by humans for humans, not technology that looks and acts human-like.

MYTH #3: Artificial Intelligence is All-Powerful

Sensational headlines like the first myth and science fiction images like the second myth can create the impression that AI can do anything. Indeed, achievements and achievements in AI are often reported, so the idea of ​​what things are not subject to AI is very vague. However, there are some things that come easily to humans but pose a significant challenge for AI.

First, AI doesn’t know how to use “common sense”. Consider the following statements: “The man went to the restaurant. He ordered a steak. He left a big tip.” What did he eat at the restaurant? Most people will immediately answer – steak. However, because AI does not have data on social norms, it will often not have enough data to understand the context. While for the AI ​​to give the correct answer it must describe in detail what usually happens at a restaurant, it does not require additional explanations for us.

Second, AI doesn’t know how to establish cause-effect relationships. If you provide the AI ​​data with the right quality and quantity, it should have no problem identifying roosters crowing at sunrise. However, artificial intelligence will not be able to determine whether rooster crow causes sunrise or vice versa.

Third, AI cannot distinguish between what is ethical and what is not. Microsoft once launched an AI-based chatbot on Twitter where anyone can turn and chat. In turn, the chatbot collected new data from these conversations to “learn” new sentence structures and speaking styles. When people found out about this, people started sending insulting messages to the chatbot en masse, such as “Hitler was right” and “I hate feminists and they all die and burn in hell” for the sake of jokes. This led to the chatbot starting to issue sexist and racist sentences because the AI ​​is not capable of assessing the ethical significance of such expressions, only copying the “jokers” language style.

Artificial intelligence is a rapidly developing technology. When you hear self-driving cars, visuals created in a single sentence, websites programmed with the help of chatbots, you inevitably begin to associate artificial intelligence with a super creature that can do anything. But if you take a more critical look at the information given in the mass media, you can finally see that artificial intelligence is just a powerful but still just a tool like a phone or a computer.

Oleksandra Vereshchak