Levels of A.I – Artificial Intelligence – Episode 3

MUSKeteer

AI is mainly categorized into three levels.
1. Artificial Narrow Intelligence
2. Artificial General Intelligence
3. Artificial Super Intelligence

We can abbreviate them as ANI, AGI and ASI

Hey guys – Welcome to Musketeer!

This is the third episode in the series Artificial Intelligence.

Before we get into the levels of A.I, let us see what is the difference between AlphaGo and AlphaGo Zero.

AlphaGo is the Google’s AI that can play the game GO. It was trained to learn from several 1000s of games played by world champions. It also improved by playing itself.

AlphaGo Zero is the next generation AI that plays GO, but it was trained by giving just the rules of the game. It started from scratch. And it’s model is much better than AlphaGo. Are games played by humans that bad? Relatively, it’s possible.

Let’s look at ANI.

The assembly line robots are programmed to do specific tasks such as painting, welding etc. Except for picking fluff from the floor, they surpass humans in speed, dexterity and accuracy. But they can not do anything else.
We have advanced ANI in several fields. Some examples of ANI are cellphones, self driving cars, AlphaGo, Siri, Chat bots, Auto Pilot etc.

Even if we code all the skills that humans are capable of, into an Android, it will be very skillful, but not generally intelligent.

Let’s move onto AGI.

In the first episode we discussed what constitutes intelligence. Let’s do a quick recap
For an entity to be intelligent, it has to be able to
1. Learn from experience
2. Adapt to new situations
3. Understand and handle abstract concepts
4. Use knowledge to manipulate one’s environment

If a machine can do this, then it can be called Artificial General Intelligence

What can AGI do?
It can do everything that a human can do. Logical reasoning, thinking, creativity, strategy, learning a new skill on its own. It can identify new sources of knowledge, it can make movies, it can do scientific research.

Some experts think that a computer can never do these. Others disagree. What about you?

In a later episode, we will discuss about the various ways humans are advancing towards AGI.

Next level of intelligence is ASI.

Apart from learning, humans have to sleep, eat, will get bored, will get tired, have fun and do everything else. But a machine that is AGI, need not do them, also it can tap into the computational capacity that far exceeding human speeds. A leading figure in A.I, Nick Bostrom, predicts that AGI can do 20000 years worth of research in a week. Given a high speed network, it can learn from every video on Youtube, every page on Wikipedia, and every course on Coursera and Udemy. And by improving its models, it become far more intelligent than AGI and humans. This stage is called ASI.

Leave your comments below about what you think it would do? What will it do with us?

Do subscribe for more and share this video with your friends.

Thank you! Have a great day!

Source