Intimidating AI…
We’ve all seen or at least heard of Sci-fi movies, TV shows, and books where the AI takes over everything. The machine frees itself from its human creators and tries to kill them, or take over the world, or something else equally scary. Hal 9000 from A Space Odyssey, Skynet from The Terminator, and I-Robot all come to mind.
…Or Not?
It could also be something less threatening, like robots becoming friendly household workers or heroes, in movies like Astro Boy. Or even movies where AI can be both good and bad, like Wall-E or The Terminator 2. And we can’t forget the various robot personalities in things like Star Wars, Star Trek, Futurama and Doctor Who.
How it Really Works
When it comes right down to it, people have always loved to imagine the possibilities inherent with all technological advances. That being said, the idea of AI in real life can be pretty intimidating to some people, whether because they’ve seen too many scary movies, or they just don’t like the idea of jobs being taken up by machines instead of humans. Most of these fears are because people don’t understand just how necessary the human element is both in creating and maintaining effective AI algorithms.
For those of you who, like me, are not techies or knowledgeable about this subject, the dictionary definition of an algorithm is a process or set of rules to be followed in calculations or other problem-solving operations, especially by a computer. To put it plainly, the algorithm sets the rules that the AI has to follow to get results.
The kind of AI we see in science fiction where a machine can understand or learn any intellectual task that a human being can does not currently exist. Researchers vary on when they believe it might become a possibility, but currently, that sort of achievement is still a long ways off. However, even if we reach that pinnacle of scientific achievement, one thing will remain the same as current AI. It has to function within its set algorithm.
Machine Learning (Or Machine Being Taught)
Of the AI types currently in use today, one of the most common subsets, and the one we focus on the most here at KarmaHub, is Machine Learning. Does it sound intimidating? Don’t worry, the AI here can’t learn anything outside of what we teach the algorithm to recognize.
For Machine Learning, there are three main components. Those components are data, the algorithm, and a model. Training data is run through an algorithm to produce a model.
To explain a little more, training data is data that has been labeled, essentially pointing out the types of patterns we want the machine algorithm to learn how to look for. But we can’t know right away if the created model is accurate, so we have to run more data through the algorithm, this time un-labeled data, to see if the algorithm learned what we wanted it to. To simplify, the data contains patterns, the algorithm detects patterns, and the model predicts patterns. It is called machine learning because the algorithm learns from the data, to create a predictive model. But this may seem abstract. Let me use an example from a movie you may have seen.
A Robot That Learns
Take the movie The Incredibles. The movie’s villain, Syndrome, builds a robot that he wants to be too powerful for regular superheroes to beat. To make this powerful robot, he has it fight many different superheroes, and it goes through several different models before he thinks it is ready to face Mr. Incredible. However, when first facing Mr. Incredible, even that model is defeated. He has to make a new one again, and that final form is the robot he uses for his evil plan. While real life AI systems aren’t used by supervillains, they do have a purpose that their creators, the data scientists, are making them for. And just like Syndrome, they have to go through many different models before they have one ready to accomplish that purpose.
Let’s break it down. Syndrome comes up with programming for his robot, to accomplish a specific task. This is like data scientists creating an algorithm. Next, Syndrome has various supers fight his robot. This is like running data through the algorithm. Then, Syndrome creates a new robot, which is like the creation of a model. This must be repeated several times, until there is a workable model that can be used in real life, like the robot that beat Mr. Incredible.
But the process doesn’t stop there. In the end of the movie, the superheroes are able to defeat the robot because of two things: its programming to obey the remote, and the use of an attack it didn’t have previous data for. This is a weakness of all AI programs today. They can only do things that they were programmed to in their algorithm, according to the patterns they were taught to recognize from the data.
That wasn’t a perfect example, but hopefully it helped to explain just how AI systems work.
AI For You
Have you ever typed in a search query and gotten no results? Asked Siri or Alexa a question and been told they don’t know the answer? That is because the training for those AI algorithms has not included data on your question yet. The AI that exists today can only search for the patterns that we teach them to recognize. Since the world is constantly changing, we also need to constantly remake the AI models to include that new information. Whether it is a presidential candidate that popped up this year, a new pop singer that rose to stardom, or the breakout of a new disease, there are constants shifts and changes that must be accounted for. That is why human involvement is essential, to keep teaching the AI all the new patterns and information that is appearing.
So what does that mean for us? Essentially, AI can only make judgements and predictions based off of past experience. And unlike in the movies, it can’t decide if something is good or bad on its own. It has to be taught that by humans. I’m not sure if this was able to relieve any fears about fictional AI, but I hope it got you interested in the real thing!
For more information on KarmaHub,
Check us out on LinkedIn at https://www.linkedin.com/company/karmahub-inc
Check us out on Facebook at https://www.facebook.com/karmahubinc
KarmaHub 