Artificial intelligence is a double-edged sword that can be used as a practical solution or as a weapon by hackers.
Artificial intelligence is now used for countless devices that we come across every single day, from changing camera quality, to face recognition and virtual assistants on our phones.
In a previous blog, we discussed some ways AI can be used for malicious attacks. In this blog, we will break down artificial intelligence and its different types, and some ways they are used in everyday life.
On May 11, 1997, an IBM computer called Deep Blue beat the world chess champion after a six-game match: two wins for IBM, one for the champion, and three draws. The match lasted for several days and received huge media coverage around the world. It was a classic battle between man vs. machine.
Deep Blue could identify the pieces on a chessboard and knew how each piece moves. It made predictions about what moves might be next for it and its opponent, Gary Kasparov. And it chose the best moves from among the possibilities.
However, Deep Blue did have not any concept of the past or any memory of what happened before. All it did was look at pieces on the chessboard as they stood and choose from possible next moves. This was one of the earliest examples of reactive AI.
The most basic type of artificial intelligence is reactive AI. Reactive AI devices always respond to identical situations in the exact same way every time, and they are not able to learn new actions, nor do they have a concept of the past or future.
Reactive AI machines are spam filters in your email that keep phishing attempts out of your inbox. They are the watch suggestions you see on Netflix or Hulu. This level of AI is the first stage of any AI system. Because of its invention, scientists were able to develop the next type of AI from this foundation.
Limited Memory AI
Google secretly launched its Self-Driving Car Project in 2009, now known as Waymo. After only a few years, its prototypes had collectively driven about 300,000 miles under computer control without a single accident. In 2014, Waymo revealed an autonomous prototype without a steering wheel, a gas pedal, or brake pedal; it was 100% autonomous.
Autonomous vehicles use limited memory AI to observe other cars’ speed and direction, helping them read the road and adjust as needed. This process for understanding and interpreting incoming data makes them safer on the roads.
Almost all present-day AI applications, from self-driving cars, chatbots, and virtual assistants are run by limited memory AI. Limited memory artificial intelligence devices have the same features as reactive AI devices, but they can look to the past to make decisions. Limited memory devices have a temporary memory. Unlike reactive machines, limited memory can learn from the past by analyzing actions given to them.
Theory of Mind AI
These next two types of AI technologies are more theoretical in nature. They are not fully fleshed out yet, but they are in development by many AI companies worldwide.
AI machines can not only form ideas about the world but also about other agents or entities in the world. This is called the “theory of mind” – the understanding that people, creatures, and objects in the world can have thoughts and emotions that affect their own behavior.
If you have ever wanted to have a meaningful conversation with an emotionally intelligent robot that looks and sounds like a human being, that is the idea behind the theory of mind AI.
This concept is being made into a reality. Humanoid robot Sophia, developed by Hanson Robotics, can recognize faces, and respond to interactions with her own facial expressions.
With this type of technology, machines will acquire decision-making that is like humans. Theory of mind AI will be able to understand and remember emotions, and then adjust behavior based on those emotions as they interact with real people.
After theory of mind AI has reached its potential, the next step of artificial intelligence is self-aware AI. When machines are aware of their own emotions, as well as the emotions of others, they could eventually have a level of consciousness and intelligence like human beings.
This type of AI will have desires, needs, and emotions as well. They will be able to make inferences (like “I’m angry because someone cut me off in traffic”) that are not possible with other types of AI.
As far as we know, self-aware AI does not fully exist. There is currently no sophisticated hardware or algorithm to support it. One of the closest things to self-aware AI was developed at Columbia University.
The group created a robot arm that learns what it is by itself; the robot has no prior knowledge, but after a day of “babbling”, the robot learned what it was, and how it functioned, from scratch. It’s a very basic version of what a newborn baby does in its crib, as it learns what it is.
Wrapping Up – Cautiously Optimistic
As developers push the limits of artificial intelligence, we should be excited about some of the possibilities of AI, but also cautious of how in-depth AI machines can get. How much progress will be made in the next decade toward theory of mind and self-aware AI? Will machines become super-intelligent AI, and surpass human intelligence?
Currently, reactive, and limited memory AI are the most common devices available. As innovative and convenient as these technologies are, they can become a crutch for everyday life. We should be mindful of the hold smart devices can have on us, and learn to limit how often we rely on them.