Will AI be able to recognize human emotions one day in the future?

Every year, machines gradually surpass humans in things we previously thought only humans could do.

Today, computers can beat us at complex tabletop games, transcribe a variety of languages, and quickly recognize almost any object.

However, it still has a shortcoming in identifying the most subtle human expressions.

Future robots may be able to make breakthroughs in sensing our emotions.

Why is this important?

Because if machines and the people who operate them can accurately sense our emotions, they can help us like never before, and even manipulate us.

But before that, let’s first discuss why something as complex as emotions can be converted into numbers, the only language that computers can understand?

Essentially, machines understand emotions the same way our brains do, through emotion recognition.

American psychologist Paul Ekman defined several universal emotions whose visual signals are the same across cultures.

For example, the picture of a smile means joy to modern urban people, and it is also true to indigenous primitive people.

According to Ekman's theory, anger, disgust, fear, pleasure, sadness, and surprise are all equally easy to recognize.

It turns out that computers’ image recognition capabilities are improving rapidly, thanks to machine learning algorithms like neural networks.

These artificial nodes mimic people's biological neurons by establishing associations and exchanging information.

To train such a network, input examples are pre-classified into different categories, such as pictures labeled as happy or sad, and fed into the system.

Then, this system network distinguishes different samples by changing the proportion of different features.

The more such training, the more accurately the algorithm can recognize new images.

This principle is similar to that of our brains, which process new stimuli based on past experiences.

Recognition algorithms are not limited to facial expressions.

Our emotions are expressed in many different ways.

For example, body language, voice intonation, heartbeat changes, complexion and skin temperature, and even the frequency of words and sentence structure in writing.

You might think that training a neural network to identify these features would be a long and complicated process, but considering the huge amount of data today and the data processing speed of modern computers, this is actually not a problem. .

From social network updates, uploaded pictures and videos, and phone recordings, to heat-sensitive security cameras and wearable physiological signal monitors, the key question is not how to get enough data, but rather should we How to use this data.

The uses of electronic emotion recognition are manifold.

For example, robots that use algorithms to recognize facial expressions can be used to help children learn or to provide companionship to lonely people.

Many social networking companies are considering using algorithms to flag special words in posts to prevent suicidal behavior.

Emotion recognition software could help treat mental illness or provide low-cost, automated psychotherapy.

Despite these benefits of emotion recognition, having our photos, communications, and physiological signals automatically scanned through a vast network is also unsettling.

What does it mean when our private information is collected by this impersonal system and then used by companies to deceive our emotions in advertisements?

What are our rights?

Current robots still need to improve in identifying subtle changes in emotion, such as identifying sarcasm and identifying the degree of emotion, and how happy or sad a person is.

However, they may finally be able to correctly identify our emotions and respond to them.

Whether they can appreciate our fear of being overly invaded is another matter.