Correcting 5 Misconceptions About AI
For the longest time, I have been interested in artificial intelligence. The idea of computers that could make decisions and think independently fascinated me. Of course, I knew that AI wasn’t quite that advanced, but the idea still captivated me. However, the barrier to learning AI was too high. It was too complex and too academic. Just looking at the mathematical notation involved made my head spin.
I eventually took the plunge and convinced my family to buy me Artificial Intelligence: A Modern Approach for Christmas, and 12 months later, I actually started reading it. I’m still reading it, and will be the first to admit I don’t understand everything that is discussed, however there are a few things that I’m realising about AI.
It’s All About Searching
Searching is a constant in AI techniques. Problems are defined and solution spaces are created. Problems can be represented in a number of ways (graphs, trees, logic knowledge bases), however in the end, it always comes back to searching.
This is important because searching is easy. Searching is something that a lot of programmers already know, even if it’s only at a most basic, brute force level. We do searching all the time. We loop through arrays searching for values, we use regular expressions to match string patterns, we retrieve records from databases. We search.
At it’s simplest, you can search using brute force searching, iterating over all combinations and permutations of solutions looking for one that satisfies the problem, but beyond that, you can involve tricky heuristics to make optimal decisions about how to search. You can have local searches, which will pick a solution space and search it for local maximums, such as hill climbing, and searches that will find global minimums, such as simulated annealing. But it’s still search. It’s still something you can do.
- It’s More Common Than You Think
AI is a hell of a lot more prevalent than most people realise. I didn’t think AI had much commercial application before I started learning about it, but luckily that didn’t dim my interest. For those who are interested, but holding back because they don’t see how it would benefit them, here’s some good news. AI techniques are used everywhere.
That international 4 city flight you booked used AI techniques. A constraint satisfaction problem solver took all the constraints about needing to be at this city by that time, flying on this airline for that much, and creating a plan for you. When Amazon recommends products you might be interested in, it calculates this with Bayesian networks and classifiers, a method of probabilistically linking a set of variables. Circuit design, product manufacturing, supply chain optimisation, all of these things use techniques that AI use. They’re not the sole domain of AI, but learning AI will cause you to learn these too.
- You Can Use It Today
Whatever you’re working on, you can probably use AI techniques in it. Even some of the more exotic sounding techniques like neural networks can be of use to you. Self healing databases? Hell yeah. Even Bayesian classifiers to catalog and categorise products, heuristic searches to mine data in databases, hierarchical task planners to plan that holiday or manage that Gantt chart.
Can you use it anywhere? No. Your simple CRUD app probably won’t benefit from a wizz-bang heuristic search. But if you’re doing anything that involves large amounts of data, interacting with people, predicting trends and recognising patterns, you can use it.
- It’s In Demand
You may have heard of the Netflix Prize. Guess what? That’s AI. Google is the biggest search engine around, and index billions upon billions of pages on the internet, and can get you relevant results to a question in a matter of seconds. That’s one hell of a big knowledge base, and one smart search algorithm. Amazon sell products all over the world, and aggressively upsell and cross sell. I get emails about related products I might like based on my wishlist and purchase history, and they’re actually pretty accurate.
Also, computer games. Enough said.
AI skills are in demand. Not huge demand, but probably more that you would have guessed. These skills are hugely profitable in the right hands, and big companies want to extract every single little morsel of useful information about your browsing, shopping, eating, travelling, viewing and reading habits in order to market to you more effectively. Now that sounds a little creepy to me, but if that doesn’t bug you, more power.
- There’s a Lot of Information Out There
AI isn’t some weirdo niche science topic. There’s actually quite a lot of information out there, once you start going down the rabbit hole. “AI: A Modern Approach” cites hundreds of papers and books. There’s thousands of websites out there on the subject. There are many academic papers that are made available for free online. There are communities, like AIGameDev, dedicated to spreading that delicious knowledge.
I think the hardest part about finding information is getting the terminology. It’s pretty dense when you first get into it, especially when talking about acyclic directed graphs, and your idea of a graph was like mine was about a year ago, namely a few bars on a 2D Cartesian axis. But once you’ve got a foot in the door of the lingo, it can become pretty accessible, and information starts becoming more bountiful. That foot in the door can be either a good, basic website, or in my case, a university level textbook designed to introduce people to AI.
AI is a big field, full of fascinating and interesting concepts and techniques, and it’s a young field that’s still full of potential. It’s not as complex or confusing as film and television would have you think. That’s not to say it’s a walk in the park, as I stated earlier, I’m probably running a 70% rate of understanding what I’m reading, but I’m managing. And if I can manage, so can you. So if you’re interested, there’s no better time to start than now.