With massive advancements in Artificial Intelligence (AI), from driverless vehicles to gaming and automated service interactions, AI is set to completely revolutionise business and life as we know it. The same can be said for AI in learning and development.
When it comes to the learning and development sector, the terms “machine learning,” “AI,” and “Deep Learning” have been making the rounds. However, they are used quite haphazardly, and are mostly viewed as interchangeable concepts. But these 3 tools are not a singular concept at all.
Here, we take a look at the differences between these tools and help you gain insight into the world of machine intelligence and learning.
At the Dartmouth Artificial Intelligence Conference in 1956, Artificial Intelligence was described:
“Every aspect of learning or any other feature of intelligence can, in principle, be so precisely described that a machine can be made to simulate it.”
Essentially, AI is the broadest way to think about computer intelligence. It refers to any kind of computer program, from a chess playing machine to a voice recognition system which interprets and responds to speech (think, Siri).
In the broadest and most simplified sense, AI can be divided into three groups. These are: Narrow AI, Artificial General Intelligence (AGI), and Super Intelligent AI.
In 1996, IBMs famous Deep Blue, beat chess champion Garry Kasparfov, while in 2016 Google’s DeepMind AlphaGo, destroyed Lee Sedol at Go – these are examples of what is known as Narrow AI. Essentially, Narrow AI is a computer/machine that is highly skilled at a specific task. This is different to Artificial General Intelligence. AGI is a human-level machine that can perform a variety of tasks.
Then there is Super Intelligent AI. Nick Bostrom describes this technology best: “This is an intellect that is much smarter than the best human brains in practically every field, including scientific creativity, general wisdom, and social skills.” Basically, Super Intelligent AI will outsmart us all.
A subfield of AI is Machine Learning. This refers to machines that take data and “learn” from this information themselves. This is the most promising tool for AI in the business world. ML has the ability to apply training and knowledge from large information to prosper at speech recognition, facial recognition, translation, object recognition, and many more tasks.
As opposed to hand-coding software programs which have specific instructions to complete a certain task, ML allows a system to recognise patterns as it learns and allows it to make certain predictions.
For example: Deep Blue, the chess playing machine, was rile-based and thus dependent on programming. This means it was not a form of ML. However, DeepMind beat the world champion by training itself.
A subset of ML, deep learning makes use of some ML techniques to solve real world problems. This is done by tapping into neural networks that simulate human decision making. Deep learning can certainly be an expensive undertaking and requires huge data to train itself. This is because there are a large number of parameters that need to be understood by an algorithm of learning, which can make a lot of false-positives.
For example: an algorithm of deep learning can be told to “learn” what an owl looks like. But it would take a lot of data set visuals and images for it to understand even the most minor details that distinguish an owl from another bird.
Again, we refer to DeepMind. This deep learning system, according to Google, worked by combining “Monte-Carlo tree search with deep neural networks that have been trained by supervised learning, from human expert games, and by reinforcement learning from games of self-play.”
In terms of business applications, deep learning can take a large amount of data, millions and millions of images, and recognise certain characteristics. Spam detection, fraud detection, image search, handwriting recognition, street view recognition, and translation are all tasks that can be done through deep learning.
To put it into perspective, deep learning networks have replaced handcrafted rule based systems at Google.