It seems like not a day goes by without seeing an article on how machines are going to take over all of our jobs. Or how a computer beat a human at Jeopardy, chess or most recently poker. Terms such as artificial intelligence (AI), machine learning and neural networks are the latest buzzwords you might be hearing in relation to this.
However there’s some misunderstanding or we could say ‘alternative facts’ as to what these terms actually mean. Clients I speak to often use the terms interchangeably or incorrectly and there’s some confusion as to what the current capabilities are. The good news though is that with a basic understanding of the concepts, you will be able to see how the technology can be applied within your own business.
I’ll attempt to briefly breakdown some basic definitions, provide real world examples of how AI is currently being used and point you in the right direction to get started exploring.
Why all of the sudden interest in AI?
A few factors have led to the sudden surge of interest and application of AI. These include;
- The rise of ‘big data’. The amount of data being generated and collected today has risen exponentially over the past few years. 90% of the world's data has been created in the last 2.5 years! Social media, the ioT(internet of things), audio and video files on sites such as YouTube as well as customer data collected on the web contribute to this. The collection and ability to analyse unstructured data, the rise of databases which can store and quickly retrieve it and the relatively cheap cost of storing it has led business to find ways to interpret this data to gain insights, reduce costs and gain a competitive advantage.
- Improved hardware. The availability and increase in power of dedicated hardware (GPUs). Dedicated and tuned for the task of processing the highly complex algorithms required for AI.
- AI Platforms. The rise of cloud based artificial intelligence and machine learning platform services is great news. Companies such as Google, IBM, Amazon and Microsoft now have platforms which everyday businesses can use to essentially ‘rent’ the horsepower and smarts needed to apply to their own data at an affordable cost.
The combination of these factors has led to a dramatically increased amount of research and experimentation in artificial intelligence.
Tell Me What It All Means
When most people think of artificial intelligence it conjures up images of robots or something like HAL 9000 from ‘2001: A Space Odyssey’. The term 'artificial intelligence' was coined in 1956 by Dartmouth Assistant Professor John McCarthy. It was defined as ‘hardware or software that exhibits behaviour which appears intelligent’. Other definitions are similar - basically the capability of a machine to imitate intelligent human behaviour.
For those with iPhones, your virtual assistant Siri is the modern day descendent of HAL 9000. Alexa, the virtual assistant in the Amazon Echo being another example. They’re still not even close to being human in terms of intelligence but they are constantly improving. Outside of speech recognition, other areas artificial intelligence is being used is for problem solving, planning and learning.
Although the concept and application of it has existed for decades, in the early days it required computer programmers to write extremely complex algorithms and supercomputers (scarce and weak by today’s standards) to crunch through them. These were difficult or too time consuming to program. Advances in computing power, software and programming methodologies and the development of new algorithms led to some of the breakthroughs which are driving artificial intelligence today.
One of these breakthroughs is known as ‘machine learning’.
I often hear people use the terms artificial intelligence and machine learning interchangeably so let’s define the difference.
Machine learning is a subset of artificial intelligence whereby machines have the ability to learn without being explicitly programmed to do so. They begin to be able to teach themselves. Artificial intelligence covers the broader scope of any method which makes machines intelligence whereas machine learning is just one (and very popular) implementation of it.
Essentially a programmer now has less work to do. Using different algorithms, a machine can be ‘trained’ to learn, make predictions and decisions by being fed more and more similar data to learn from. The algorithms themselves are generally statistical or predictive analysis spotting patterns in data without being told by a human where to look.
For example, your Facebook News Feed uses machine learning to help the information you see (and accordingly serve you relevant ads). Machine learning is the process Netflix uses to give you recommendations of TV shows it thinks you will like. The more TV shows you watch, the more it learns and adjusts accordingly. When a Google search auto-completes a search term for you or corrects a typo... that's machine learning at work.
'Deep learning' is a term you will be hearing more of in regards to AI. It is sometimes mistakenly used interchangeably with machine learning but is really a subset of it. Deep learning can be thought of as the cutting edge of machine learning. This is where some of the most exciting advances in AI are currently being made today. Big players such as Google, Facebook and Microsoft have advanced this technology with great success in recent years.
With deep learning a machine is modelled to act like the human brain in order to learn and teach itself. To do this, data is fed through what are known as neural networks to continuously process and reprocess the information to make decisions about it, much in the same way the human brain does.
As an example and to show the difference between this and traditional machine learning, take the example of training a machine to recognise photos of dogs. With machine learning you would need to write an algorithm to include all of the features that make up a dog such as the number of legs, fur, size, colour, different breeds etc. But dogs come in many shapes, sizes and breeds and it’s almost impossible to write an algorithm to cover all of these differences.
With deep learning, the machine can define the features that make up a dog itself without being told, start to classify differences in breeds and begin to correct itself and improve it’s predictions made from previous mistakes.
Doing this requires huge data sets and computer power to process. A great example of this is Google Translate. I highly recommend this New York Times article - a fascinating read on how deep learning was applied to greatly improve the accuracy of Google Translate. It also gives great insight into how deep learning was developed.
Some of the applications of deep learning being used today are to;
- recognise faces, objects and landmarks in photos.
- analyse documents and audio conversations to detect sentiment and emotion.
- detect tumours or other abnormalities in medical images.
- build self driving cars.
How Can My Business Use Artificial Intelligence?
As complex and expensive this might sound, the good news is that companies such as Google, Amazon, IBM and Microsoft now let you tap into their artificial intelligence engines. For complex applications, the need for developers and a data scientist might still be required. However, these platforms also offer ready made services which you can feed your own business data into.
Some useful places to start are;
Need to build an app with image recognition capabilities or which converts speech to text? Not only is this possible but relatively simple. How about a customer service chatbot on your website or an application which scans customer report requests, automatically classifies them and even answers simple requests itself. It’s all possible.
I also highly recommend running some of your business data (sales, finance, HR etc) into a cognitive data analytics platform such as IBM Watson Analytics. It will automatically generate data visualisations, find patterns and trends in your data and make predictions based on it. All at a very affordable cost.
Businesses trying to make sense and gain insights into their data, who want to save costs by automating repetitive tasks or who are looking for a competitive advantage should be exploring the ways artificial intelligence can help them. There’s never been a better or more affordable time to start integrating AI into your business strategy.