Artificial Intelligence: Taking a Closer Look

Technology itself, as Wikipedia states it, is the application of scientific knowledge for practical purposes. As soon as we are able to invent tech that supports us in one way, we use it to try out something new. We wanted to fly, so we created planes to glide like birds over the oceans. We drive cars and trains that are faster than jaguars through cities and landscapes sculpted with our hands and brains. The last 30 years have yielded millions of new patents – all new inventions that have jumpstarted our state of technology to where we are now: checking the time on the same device that connects us to the whole world.

A Quick History Lesson

One rather new technology provides us with 'brainpower', to be more specific thinking, learning and knowing. The keyword is Artificial intelligence. Well, I have to admit, ‘new’ is the wrong word. AI has been around since the 1950s, or at least the theory of it. McCulloch and Pitts ‘A logical calculus of ideas immanent in nervous activity’ was one of the first papers to appear on the topic of AI in 1943. Followed by Turing himself in 1950 with his paper on ‘Computing Machinery and Intelligence’ (which, by the way, provided the name for the movie ‘The Imitation Game’).

In 1956, the Term ‘Artificial Intelligence’ was adopted by the Dartmouth Workshop and it may as well have been the starting point of AI as a field of research. A couple of names that were attending (and you should know by heart when interested in AI):

• Dr. Marvin Minsky, pioneer in research of AI
• Dr. Claude Shannon, father of information theory
• Dr. Julian Bigelow, father of lisp

The next ten years were filled with new discoveries and logical algorithms. From the mid 60s to the late 70s, AI and Machine Learning disappeared almost entirely. At that time, computing power was the biggest issue and forced researchers to abandon the idea. Since approximately 2010, this problem is solving itself at an incredible velocity and AI (mainly Machine Learning) is experiencing a revival.

Don’t get me wrong here, a lot happened between 1970 and 2010. Knowledge based systems as well as expert systems in the industry boomed in the 80s.

Building Intelligence

Like many other advancements in technology, nature can be a good inspiration/example of how an AI could be developed, one of the best examples being neural networks. Sparked by different, newer technologies – for example the MRI, that allows us to see inside of our own brain – AI-research is striding forward like never before.

To build a human-like intelligence or a more advanced intelligence that we could not possible imagine at the moment, we need to know a few basics.

From early on, it was clear that building a full system (like the brain) would not be something that we are going to be capable of any time soon. So, what do we engineers usually do? Break down the problems, generalize and solve them in small parts. These small problem solvers are called agents. An agent basically maps input or converts histories to outputs. To make an agent intelligent, we often work with rational agents. These agents maximize the expected value of their actions, given a specific task environment.

As of today, individual agents are really good at specific tasks like image recognition, speech recognition and even producing small sport-reports. The challenge lies in connecting these agents into one single thinking, learning machine.

Stay tuned for more. In part 2, we’ll look at a few agents, systems that learn by themselves and where we might be at in a couple of years.