Let’s talk about artificial intelligence – or AI. What was once the stuff of science fiction movies, speculative novels, and the dreams of particularly ambitious inventors, is now very much a reality. Everything from the predictive text on your cellphone and the results supplied by your search engine of choice, to the traffic signals you negotiate on your way to work; all of this utilizes AI to some degree.
Yes; artificial intelligence has well and truly landed, and it won’t be going away any time soon.
It is a shame, then, that so few of us fully understand such an ubiquitous technology. We use it, we enjoy its benefits, we probably grasp it to some extent; but few of us can say we deeply understand it.
This is natural – it’s a pretty complex topic after all – which is why we need illustrative, insightful visuals to help us get to grips with it; visuals like those produced by Daniel Smilkov.
We know all about learning from data and gaining vital insight from the facts and figures we receive; we also know all about interpreting this data and applying it to great effect within business, but what about our systems? How can computer systems and pieces of technology learn to understand the data they handle?
This is where neural networks are helping data scientists and engineers to make great advances in artificial intelligence and machine learning. These networks enable software to interpret and gain insight into datasets, and exist at the cutting edge of a rapidly developing field of technology. But what are they, and how do they work?
As you might expect, the developers of neural networks took inspiration from the most powerful biological computer on earth; the human brain. Just like the human brain uses neurons – cells which communicate with one another via electrical and chemical signals which pass backwards and forwards – neural networks mimic this with their own ‘digital neurons’.
When the network is in operation, these digital neurons send each other messages and record the response. By setting the network to work on an objective, the neurons use this communication to build up a picture of the various routes towards this objective, discovering which ones work and which ones do not via a process of trial and error. The network ‘remembers’ which routes led to failure and eliminates them from future iterations of the task.
Conversely, winning connections are strengthened and ‘remembered’. The network learns.
Smart Visual Explanations
This is a very basic explanation of a fascinating and enormously complex field. However, it is worth trying to get a little closer – taking cues from those intelligent networks and developing our understanding ever further.
Machine learning using neural networks is – after all – a huge part of our lives. Take a look at the giants of the modern tech industry. Chances are, you own one or two of their products. Each of those firms is investing time and capital in machine learning tools and technologies, as are your favorite search engines and social networks. The prevalence of this technology is only going to increase as it improves and as organizations learn how to apply it to their own high tech products. In other words; get used to these networks, and take steps towards understanding them, as they aren’t going away!
Daniel Smilkov’s visuals represent the ideal way to gain this understanding. Working with Shan Carter from Google, Smilkov created an interactive map of a neural network, enabling users to adjust different parameters, swap values, and just generally tinker with their creation. With the network complete, the user needs to simply flick a switch and watch how the whole thing unfolds.
There is a world of difference between being told about a complex process – either in an online article like this one, or in a lecture theatre at college – and actually seeing it in action; actually witnessing and influencing its different outcomes. This is an instance of active learning; something which is far more beneficial to us and to our understanding than a simple, passive event.
I’m not the only one to have been impressed by Smilkov and Shan’s data visualization wizardry. The work of the two visualizers has been causing something of a storm online, and not just among those already operating in the industry. Have a search for “Smilkov, Shan and neural networks” yourself, try your hand at creating a smart network, and witness the buzz currently being generated by fantastic visualization.
The attention surrounding the work of Smilkov and Shan is a testament to the power of this medium, and the advantages it provides to those seeking to hone their understanding of a fascinating field.
Data Vis and Complex Knowledge Transmission
Smilkov and Shan have created something truly special here, but this is just the tip of the iceberg. By distilling the details of such a complex field of study into a smartly interactive piece of visualization, they have turned the visual medium into the key which unlocks a deeper understanding.
We have seen how visualizations have been utilized to great effect in the education sector. Everyone from primary school teachers through college professors, right up to corporate trainers and adult education specialists use visualizations to break down points, to provide students with valuable insight, and to develop psychological associations between process and definition.
Of course, the transition of complex knowledge is not limited to the seminar rooms and lecture theaters of our nation’s universities; it is a concept which is used in business almost every single day.
Business is a dynamic discipline – one which requires a degree of agility and flexibility if we are to succeed. This means taking onboard new pieces of information – often relating to complex and sophisticated concepts – quickly and efficiently. What’s more, we need to be able to transmit this knowledge to our teams in a similarly timely manner.
Ideas which previously seemed difficult to grasp can be transformed using a high quality visualization, enabling us to roll out new procedures, new processes and new concepts within business. Active learning is the future for our organizations; and data visualization is bringing that future to us.
Expand your intelligence capabilities as your business grows. Find out how SAP BusinessObjects Lumira can help.
Image via Playground.Tensorflow.org