The Story of AI

If you eyeballed this article expecting the typical ‘AI is bad and is going to steal our jobs’ doctrine, or hoped for more tales of a two thousand word essay on the causes of the US Civil War being written in seconds, this article is not for you. This is the story of how artificial intelligence came to be as famed as it is today.

AI is now one of the biggest buzzwords around, but it really doesn’t mean much when it comes down to it. For something to classify as ‘AI’, all it must be is man-made and intelligent, the latter being the much more nuanced term. Much like organic life, intelligence can be a simple condition; like “swim towards the light” for algae, or the complex deductive reasoning needed to understand anything such as understanding television puns. However, for code, that could be anything from the complex black box algorithms used to recommend content on social media to chatbots that talk to Year 7s. When referring to AI, what people generally mean is ‘Machine Learning (ML)’. ML is the ability for a program to adapt and improve as new data is fed in. This, as a lot of programming related topics tend to have been, started with Alan Turing.

In 1947, Turing gave the earliest public lecture mentioning machine learning, talking about the possibility of ‘letting the machine alter its own instructions’ with references to chess; a complex game with a number of possible unique games slightly larger than three, but not by much. Chess presents an interesting challenge as its complexity makes it impossible to “brute force” which, in this context, means expending a large amount of computing power, time and energy to calculate all possible futures, like Dr Strange but without the magic.

So, how do you solve such an impossible game?

You give up.

Or you could employ the art of heuristic analysis, taking the mentality of the average CS student finding an error in their coursework and ignoring it if it’s not too major. However, in return for sometimes making a wrong move, chess computers need to check much fewer possibilities. In the example of chess, you would not want to lose your queen. Therefore, it is unlikely that any move which loses you your queen with no return is bad. In this spirit, bots may safely ignore any move sacrificing the queen, causing mistakes only on the rare occasion that the best play was to forfeit the piece.

This, combined with the exponential advances in microprocessor power (roughly doubling every 50 years with Moore’s Law) allowed IBM to make Deep Blue exactly half a century later in 1997. Since then, the field of AI has advanced significantly when trying to solve games. So far, that making ML programs to beat games can be the subject of a quirky YouTube channel. However, the main applications that have come under the attention of the public eye are generative AI's, namely ChatGPT and DallE.