As you have seen, artificial intelligence is already a part of our daily lives, and scientists are finding many more ways of using it. With AlphaGo, the computer program that can autonomously learn how to play board games, Sophia, humanoid robots by Hanson Robotics, autonomous cars, and ChatGPT, the news is full of impressive examples of artificial intelligence.
These innovations and, more generally, future possibilities for this technology have been stimulating artists’ imaginations for many years. Think of Frankenstein, the Terminator, Wall-E, and the Black Mirror series—to name a few: our culture is replete with depictions of hypothetical future technologies and augmented humans.
But let’s take a moment to separate myth from reality, to uncover AI's true potential.
Debunk Common Myths Surrounding AI
Myth No. 1: "AI Programs Are Smarter Than Humans."
Artificial intelligence programs can accomplish amazing feats, such as defeating the world’s best Go players, identifying rare species in videos, and even bluffing their way through poker matches.
However, it’s essential to recognize that these programs are not copies of the human brain but instead inspired by it.
Current artificial intelligence programs are not that intelligent. Above all, they are incredibly specialized and are therefore very efficient at performing particular tasks.
As Yann LeCun put it:
"The most intelligent computer systems today have less common sense than your cat!" 🤭
For example, if I were to say, "Rose left the building with her suitcase."
Since you are not a robot, you possess common sense. Therefore, you have gleaned a lot of underlying information from this simple sentence:
You know that Rose is a person and not a flower.
You understand that Rose is no longer in the building.
You have even formed the hypothesis that Rose is going on a trip since she has a suitcase.
Machines haven't yet mastered the ability to apply judgment.
Myth No. 2: "AI Functions Just Like the Human Brain."
The goal of artificial intelligence is to solve complex problems that would typically require human intelligence. For example, it might involve visual perception or language recognition.
The development of neural networks (and thus deep learning) attempted to mimic the human brain's functions —and these neural networks make it possible to dialogue naturally with a machine.
As you saw above, current AI programs are very specialized.
Intelligence can be measured in different ways. When it comes to quick mathematical calculations or large memory capacities, artificial intelligence outperforms humans. However, there are other capabilities that AI will not possess for a long time. For example, AI does not have the capacity for emotion, empathy, or humor, although it can appear to portray these traits. 😏
Myth No. 3: "AI Programs Are Conscious and Might Have Feelings."
Before the advent of AI as a scientific discipline, artists imagined creatures capable of consciousness and emotion. This idea fuels countless works of fantasy and science fiction.
Pixar's Wall-E robot appears to express many feelings. In science fiction, humanoid robots often possess intelligence and emotion.
Outside the realm of science fiction, some smart speakers can now express emotion. Amazon Alexa can respond using intonations to show excitement, disappointment, or disapproval, which can be useful as an artificial companion (i.e., to relieve an older adult's solitude, extend a warmer welcome to hospital patients, etc.). Microsoft’s Bing chatbot hit the headlines for generating aggressive and hurtful responses in conversations with some users. In the next part of the course, we'll see that while this behavior might express powerful emotions such as anger, it cannot be compared to human emotions.
Myth No. 4: "AI Programs Could Take Over."
It’s important to understand that AI models' behavior and purpose depend on how humans design them.
Some big names in science and technology, including Stephen Hawking, Bill Gates, and Elon Musk, have expressed concern about a super-intelligence, artificial intelligence that could develop beyond humans' ability to control it. In cases where AI models are assigned malicious purposes, they could develop into phenomena beyond human control.
These fears highlight how difficult it is to correctly define what we want our AI systems to achieve and where to set up failsafe controls. We’ll explore this in the chapter on safety. If we take an AI system capable of carrying out many different tasks and don’t define its purpose properly, these types of incidents could become a reality.
But still, we’re not talking about AI systems that can spontaneously develop their intentions and purposes, as in the film “iRobot.”
While AI systems are inspired by the human brain, they are not a copy of our brains.
AI products don’t experience emotions as humans do. They are only simulations of emotions.
AI systems don’t possess free will. Their behavior and purpose depend on the way humans design them.
Now that you have learned more about what artificial intelligence covers, test your knowledge by answering a few questions!