The History of Artificial Intelligence
These days AI is all around us: every time you use Google Search, Translate, or when you watch something on Netflix you’re interacting with AI. Every like or comment on Instagram, Facebook, or Twitter triggers algorithms to do their magic stuff. Did you know that some modern phones even have a dedicated ‘neural engine’ within their SoC to amplify artificial intelligence? But the field of artificial intelligence has been around us for longer than you might think.
Early Artificial Intelligence
In 1950 Alan Turing developed the Turing test, to test a machine’s ability to exhibit intelligent behavior. Will a computer ever be able to sufficiently imitate a human to the point where someone cannot tell the difference between human and machine? Despite major advances in AI, no computer has ever passed the Turing test yet.
In 1959 Arthur Samuel published an algorithm for a checkers program using machine learning. It was self-learning because it remembered every position it had already seen, along with the terminal value of the reward function. Therefore, the machine was able to effectively extend the search depth at each of these positions. He played thousands of games to train his model. He continued to work on checkers until the mid-1970s, at which point his program achieved sufficient skill to challenge a respectable amateur.
The field of AI had its ups and downs, but in 1980 ‘expert systems’ became booming. These systems were programmed to mimic human experts. Facts and rules were hard-coded into the machines, usually represented in the form of ‘if-this-then-that’. These systems could represent knowledge of different subjects. Expert systems were among the first truly successful forms of AI.
Major Breakthrough: Deep Blue
Let’s fast forward to the 1990s: AI solutions had successes in speech recognition, medical diagnosis, and many other areas. IBM had been working on a chess computer for years, called Deep Blue. In February 1996 a six-game match began between Garry Kasparov, considered one of the greatest players in the history of chess, and Deep Blue. Although Deep Blue was capable of evaluating 100 million different chess positions per second, the human chess master triumphed over Deep Blue in the sixth game and took the match, with a final score of 4–2.
Did you know?
The name Deep Blue is a combination of its former name ‘Deep Thought’ and ‘Big Blue’, IBM’s nickname.
The IBM team kept working on upgrading Deep Blue and improved the computer to examine 200 million different chess positions per second. During a rematch one year later, Deep Blue won by 3½–2½. IBM’s computer stunned the world by becoming the first machine to beat a reigning world chess champion under tournament conditions.
Rise of Deep learning
In 2009 the ImageNet database was presented at the CVPR conference. The database now includes more than 14 million images that have been tagged by humans to indicate what objects are pictured. Since 2010, the ImageNet project runs an annual software contest, the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), where software programs compete to correctly classify and detect objects and scenes.
In 2012 during the ImageNet challenge, a deep learning model using a convolutional neural network called AlexNet achieved a top-five error of 15.3%, blowing out the competition that was more than 10.8 percent points lower than that of the runner up. So from there, the practical applications for deep learning began to take off.
Stanford researchers made similar breakthroughs in 2014 when creating computer vision algorithms that can describe photos. Until then, computer vision had largely been limited to recognizing individual objects. But this algorithm teaches itself to identify entire scenes: a group of young men playing frisbee, for example, or a herd of elephants marching on a grassy plain. So, rather than just saying this is a human or animal, it’s able to give a bit of a story behind an actual picture.
Major Breakthrough: Project Debater
A more recent breakthrough is IBM’s Project Debater in 2018. It’s an AI system that can debate with humans on complex topics. It’s able to debate topics it never trained on before and can comprehend human speech in order to construct a meaningful rebuttal. On February 11, 2019, Project Debater debated Harish Natarajan, who holds the world record in number of debate competition victories. The team behind Project Debater prepared 7 months for this challenge specifically. Nothing, except for the first sentence used to greet the human debater, was pre-programmed or pre-scripted. Both sides debated the topic “We should subsidize preschool”, which took place in San Francisco, in front of a live audience of around 800 people, and was hosted by Intelligence Squared and moderated by John Donvan. It’s fascinating to watch and would highly recommend looking into it.
I only covered a few of the many major historical highlights within the field of artificial intelligence. But there’s more, so if I got you interested please do some research and share it!
The future of AI
How about the future? Experts think that AI will outperform many of the human tasks within the next 45 years. In the near future we’ll see machines making medical diagnoses, designing buildings, composing music and self-driving cars. It will fundamentally change our lives in many different ways, but also raises concerns regarding liability and privacy. Let’s see how this develops over time…
