Human neurons operate at about 200 Hz, while axons may carry action potentials at the approximate speed of 120 m/s. What about modern microprocessors that emulate human brain activity? Well, they perform at a one million times faster speed with the ability of computer signals to travel near the speed of light. Do you still think that AI take-over is not possible?
Fortunately, not yet. But just in case humans are not careful enough in further AI development, it will be quite a probable future scenario. And even though an age-old “humans vs. robots” rivalry won’t likely materialize soon, the take-over is around the corner in a more pragmatic sense. AI is already capable of penetrating most industries and replacing humans in some occupations.
There’s nothing new in AI predicting stocks prices or improving weather forecasting, but what about the AI predicting the date of your death or defining your personality? That’s grand. Let’s see how exactly it can do this and reveal five more human things AI can already predict better than humans.
- Chances of Dying
According to the recent news published by New Scientist, AI can predict the person’s chances of dying within a year by analyzing electrocardiogram (ECG) results. It can do it even if the ECG is absolutely normal and the person is healthy.
AI analyzed over 1.77 million ECG results from nearly 400,000 people using two machine learning models: raw data processing and processing of the ECG data combined with human-derived measures and common disease patterns. And guess what? The predictions based on raw ECG data turned out to be more precise.
- Signs of Aging
Skilled doctors may find out that someone is sick just by looking at them and comparing how they look to their age. If someone looks older than expected, it’s a wake-up call to get tested. But unlike the chronological age, the so-called biological age is extremely hard to determine, as it requires an awful ton of factors to be taken into account.
This study, however, indicates that the AI improvements open up new capabilities for revealing the signs of aging in an individual. AI, unlike humans, can quickly analyze all biomarkers gathered from smartphones and wearables — if the person gives consent on personal data collection. These predictions, scientists say, will help the industry understand the process of aging better and deduce the formula of healthy and happy aging.
This, combined with other breakthrough technologies like anti-aging drugs, can be a major factor influencing lives of future generations, and AI just might be a part of it all.
- Addiction Risk
This one is probably the most surprising. Here, AI doesn’t use biomarkers or results of some medical tests to find out who is prone to substance addiction and who is not. Instead, algorithms analyze… Instagram accounts. Images that a user posts on social networks and captions below them can give some clues about their personality and lifestyle preferences.
A group of enthusiasts created a machine learning model that can extract specific features from both images and texts. To educate the AI to detect these features, they crowdsourced thousands of participants to share their Instagram education, analyzed their profiles and defined the possible signs of people getting addicted to alcohol, tobacco, or drugs. The AI was specifically successful in detecting alcohol addiction, scoring a 72.4% accuracy.
Besides substance abuse, AI can make assumptions about someone’s chances to get addicted to specific behavior. Let’s take a look at gambling as an example. Most gamblers, just like alcohol consumers, engage in this activity only casually, meanwhile, some percentage of them become desperately addicted and often go bankrupt.
Fortunately, both offline and online casinos gather a tremendous amount of data about their visitors and users. According to this report, such data along with data about the typical causes of gambling addiction and psychological traits of addicts may help AI algorithms to predict who might become a problem gambler with a 60% probability. This probability can be increased up to 90% if more data is gathered during detailed interviews with gamblers.
- Mood
Your next revelation of the day: AI can know how you feel today from the way you walk. Surely, we as humans can even sense the mood of a stranger from their facial expressions and movements. But how exactly does AI, a soulless algorithm that lacks empathy, do it?
Researchers from the University of Chapel Heal invented a machine learning method that may help AI recognize the human mood from the gait alone. They picked four different emotions: happy, sad, angry, and neutral and recorded several videos to compile a 3D model of poses belonging to humans in a specific mood. The distinctive features that matter here are shoulder posture, head tilt angle, the distance between consecutive steps, “movement expansion”, and some more.
The result? AI managed to predict a human mood with 80.07% accuracy, according to preliminary experiments.
- Personality
Another thing we are used to is defining someone’s personality from how they look. We pay attention to all attributes like clothing, gait, haircut, and others but the most definitive one is face. However, studies that involved human raters in assessing specific personality traits just by looking at the photographs got inconsistent results.
Newer research revealed that AI could outperform humans in this guessing game. 12 thousand participants took part in the experiment, where they contributed 31 thousand selfies along with a self-report survey regarding their personality. Then, neural networks processed the data to weed out images with emotional expressions, photos of celebrities, etc.
Researchers then taught AI to decompose each image into 128 invariant features and give some personality verdict based on this data. A reported accuracy of the AI comprised 58%, which is more than a previously defined accuracy of an average human being.
Such news may plunge us into a profound reflection on the human purpose. Can it happen that we humans will be needed only to educate and program AI to do 99% of the job? And what if AI will predict all our life patterns (and finally conquer humanity, for sure)? Breathe in, breath out. There are still more things AI can’t predict than it actually can.
For example, even having a ton of data under the belt, AI can’t predict how a child’s life will turn out. Hundreds of researchers tried to use 15 years of data to figure out the future of a specific family or an individual, mostly to predict employment and crime rates. However, none of these attempts succeeded: AI hasn’t shown any decent accuracy.
