AI Versus The Human Brain
AI systems are often compared to the human brain, even though they have almost nothing in common. To achieve artificial general intelligence (AGI), we tend to look to the only example of general intelligence available for humans to study: the human brain.
This approach has led to modern artificial intelligence (AI) often being presented as working like your brain. It is more accurate, though, to view modern AI techniques such as machine learning (ML) as powerful statistical methods. The system analyzes quantities of tagged data looking for commonalities and correlations. One problem with most AI systems is that even when a system is working, we often have no idea why. And when a system doesn’t work, we can’t identify and correct the problem.
Humans can solve a variety of problems and learn to solve ones we haven’t encountered before. In a similar vein, the fictional computer HAL, from the movie 2001 A Space Odyssey, could solve problems generally and continuously tackle new problems based on learned information just like humans do. Today’s AI has not yet reached this level.
One issue that today’s AI has with demonstrating intelligence is exhibiting common sense. Current AI, for example, lacks the common-sense knowledge to recognize that:
• Physical objects exist in a 3-D reality and persist even when you can’t see them.
• Objects have numerous properties and are subject to physical laws, such as gravity.
• Time passes and imposes a certain order to actions in the environment.
• Objects in motion follow generally predictable paths such as falling, rolling, etc.
• Causes can predictably lead to effects.
• Actions that a person (or AGI) takes can influence the future, which may impact the person.
Driving a car provides a prime example of how common sense is inherent in general intelligence. Let’s say you’re driving down a residential street. If you see a child in a yard playing with a ball, you can predict the ball’s motion, followed by the child’s motion. As a result, even if the child goes out of sight behind a parked vehicle, you still know that the child exists and could emerge in the car’s path without warning.
Similarly, an AGI vehicle encountering similar circumstances should be able to project what may occur and take the proper precautions through its own intelligence. But while today’s self-driving vehicles can be programmed to recognize and predict balls, children and parked cars, their AI would likely not apply to unpredictable events, such as the ball rolling into the street or one of the children disappearing from view behind a parked car (and then perhaps running into the street to chase that rolling ball).
