Artificial intelligence is a fascinating topic and is quickly becoming integrated into every piece of technology from Smart Phones to refrigerators.
While not everyone cares about the inner details of AI I really recommend reading this two-part article I will mention in a bit. The impact AI will have on our species will be larger than you can possibly dream of.
When most people think of AI, they think iRobot and Skynet, but as you can see from recent technology announcements, it will be integrated into every aspect of our lives. AI will be used to solve problems you never knew existed. In fact, many instances of AI you won't even recognize as AI.
To understand the true meaning of this, you need to understand a little more about the levels of AI.
Levels of AI
- Artificial Narrow Intelligence (ANI)
- Artificial General Intelligence (AGI)
- Artificial Super Intelligence (ASI)
Artificial Narrow Intelligence (ANI)
This is your typical AI that plays chess and your typical learning thermostat. It is built for one purpose and it is very good at that one thing. If you were to take a chess AI and have it play checkers, it wouldn't have a clue what to do.
Almost all AI we have today fall under this category. It is the easiest to built and solves most problems it is built for reasonably well. ANI is typically trained by throwing lots of data at a model and having it learn the difference between the solution and the wrong answers.
Artificial General Intelligence (AGI)
We are already now in the future. We have not solved this level of AI.
While ANI is very focused on one task, AGI is able to adapt to almost any task thrown at it. This is the type of AI you typically see in movies. We have yet to figure this out due to how our current AI solutions work today.
When you build an AI you need to train it on hundreds of thousands of samples or even millions to get a reasonable (80%+) success rate. The more training you give it, the more accurate it can become.
The problem with this process is you are only training it to solve one problem. We currently don't have a way to train a robot to be good at everything and adapt to problems like a human can. Like most computer programs, an AI is only as good as the developer makes it.
Artificial Super Intelligence (ASI)
This is likely the end of the world depending on the motivation of the person developing it. ASI is the point where artificial intelligence is smarter than we are at everything. This could be the difference between a Harvard graduate and a typical college or could be the difference between Steven Hawking and a potato.
The thing is, once we get to this level, the growth will become exponential. The smarter you are, the faster you learn. As AI becomes smarter, it will be able to get even smarter faster. If for example, we develop an AI that is five times smarter than us today, in a month it may become ten times smarter than us, and in a year a thousand. At these, levels the most complex problems we face may become trivial. Solving aging, cancer, and AIDS may become as simple as solving a beginner Sudoku.
The problem we face is how the AI is bound. If you watched Robocop, you will remember the robot had certain Prime Directives to follow.
- Serve the Public Trust
- Protect the innocent
- Uphold the law
- Any attempt to arrest a senior officer of OCP results in shutdown
These directives sound well and good and are obviously well-thought-out and appear to be for the greater good. What happens when a superior intelligence realizes humanity is the biggest threat to the innocent?
The two-part article I am going to recommend goes into a scenario like that with a robot designed to be the best Handwriting robot in the world. This simple and apparently harmless goal becomes far more dangerous than you will imagine.
There is a famous two-part article on The AI Revolution: The Road to Superintelligence by Tim Urban. I recommend this article all the time to people interested in AI and it is a really great read. It varies in complexity throughout the article, you can skim over the more complicated portions you don't feel comfortable with. If you read nothing at all, I recommend reading the story about Turry The Handwriting Robot in Part 2 to get a good idea of where I'm going with this.