“That guy is so good at math. He is like a walking calculator!”-ever heard that one before? Or…”She is so smart, she has a computer for a brain.” Although these statements might seem complimentary, the implication would really be calling someone dumb. Because computers are still dumb. Sure, they can compute some of the most complex algorithms in a matter of seconds, but they have no common sense. But.what about Artificial Intelligence? When those two words are thrown together (or even two letters-AI), so much is triggered in our brains. Super-genius computers, the Terminator, robots taking over the world. Slow down. Ok, so what exactly is artificial intelligence, and are we there yet?
Today we are more accustomed to the process of machine-learning. Machine Learning is allowing computers to learn code on its own, rather than us explicitly programming what it should do. One very common example is having a computer know what a cat looks like. Well, we could write some explicit code like that. You might even start hacking away with a bunch of nested “if” statements, “If animal has whiskers, if animal purrs…etc.” Not only is this annoying and tedious, but it is very hard to wrap all the cats out there in one class and expect the computer to visualize what you are talking about. Think about how different a kitten is to a cheetah.
Machine learning to the rescue. Instead of us having to explicitly type to the computers, we can show the computer thousands of different images of cats, let it pick up on themes and similarities, then maybe show it pictures of similar animals that are not cats (maybe some dogs), so that the computer can process the differences. Over time, the computer will eventually be able to have the “wisdom” to be able to spot a cat. Fantastic. It actually is kosher to technically label this as “Artificial Intelligence”, as machine learning is a subcategory of AI.
Wait, but what about Terminator, Wall-E, the amazing robot salesman from the “Adventures of Jimmy Neutron?” I am sorry to disappoint you, but we are not exactly there yet. This concept, of computers being just as smart as a human, is a different branch of AI, known as artificial general intelligence (AGI).
Alan Turing, one of the fathers of modern technology and programming, proposed a reasonable system to test for artificial general intelligence. If a robot through answering a series of questions, can convince a blind judge that it is a human at least 30% of the time, it would have passed the Turing test. No computer has passed such a test. Steve Wozniak proposed a coffee test to measure true artificial general intelligence, whereby a computer would have to make coffee without any other instruction. It would need to figure out what items to use, where to find them, and then perform the task as a human would. Again, no robot has passed such a test.
So, how far away are we from AGI? Top researchers were interviewed, and although the range was quite wide with few, straight forward answers, the average estimate was about 2099–81 years from now. Basically, we do not know.
So, if you are a coder, get excited! You can be a part of the race to create AGI! What I implore you to do is invest in working on the morality of these machines, as what many researches are also considering.
Sources:
https://medium.all-turtles.com/the-turing-test-is-tired-its-time-for-ai-to-move-on-8d636ce50795
