From Metropolis to the Terminator to this year’s Ex Machina, technology capable of thought has been an enduring concept in dystopian science fiction. The technophobia that drives these story lines has seen a resurgence of late. Sensationalist news stories about artificial intelligence – including those that discuss the increasingly buzzworthy field of machine learning – are largely to blame.
But do the critics have a point? Are learning and thinking processes sacred? Perhaps more to the point, is it dangerous to equip machines with the capacity to learn (a capacity that, in human beings, has lead to acts of destruction as well as quantum leaps in innovation)? Most people who envision doomsday scenarios brought on by killer robots have a poor understanding of machine learning (I say “most” because this group includes some highly-knowledgeable scientists and tech leaders).
If you haven’t made up your mind either way, you may want to learn a bit more about machine-learning, how it’s being used now, and the magnitude of its potential benefits.
It’s no wonder people are wary of the digital world. It seems every week a new profession is pronounced dead at the hands of technology (not “transformed”, as is almost always the case, but “dead”). This type of anxiety hasn’t reached such heights since the industrial revolution, when the mechanization of processes formerly carried out by hand changed the world irrevocably.
For obvious reasons, artificial intelligence (or AI) has elicited the most technophobic sentiments. Fear has reached beyond expected circles, reaching the likes of Elon Musk, Bill Gates, and Stephen Hawking. As the area of AI that “gives computers the ability to learn without being explicitly programmed”, machine learning is sure to be hotly debated over the course of the next few years.
Having said all of that, most companies carrying out research in these fields aren’t impacted by the negative press – at least not significantly. Google and Facebook are two of the very powerful entities investing in machine learning and artificial intelligence more generally. The benefits these companies experience will likely offset any negative public responses they receive. As far as I know, nobody here at Gnowit has encountered any anxiety with regards to the company’s technology, and we use machine learning to improve the automatic monitoring and filtering that occurs during the web-intelligence collection process. Nonetheless, at the current time, technophobia is so present in pop culture that it’s worth exploring.
Are you disturbed by the fact that IBM’s Watson (a cognitive computer) beat three human contestants at jeopardy? Perhaps you’re weirded out by the everyday applications of machine learning, like the Amazon recommendation engine that seems to know you as well as your best friend. Either way, a brief explanation of the power behind these applications may set your mind at ease.
There are some good descriptions of machine learning out there, but one of the data scientists here at Gnowit did a particularly job of explaining it to me. “Machine-learning systems use the data they encounter to improve themselves,” says Dr. Andrew Droll, who has been working on a major machine- learning project with the company. “They make predictions on the basis of statistics. When they’re wrong, they incorporate their incorrect guess into these statistics, improving future responses”. For human beings, “experiences” is a more appropriate word than “statistics”. Still, these self-improvement processes bear some serious similarities.
If you find this worrying, remember: computers can carry out a process akin to human decision making, but when you get right down to it, they’re simply following the instructions entered by their programmers. As Dr. Droll puts it,“machine-learning tools can make highly subtle and complex distinctions. But when you get right down to it, everything is still all in the code.”
How machine learning is changing things
So, what’s the point of programming computers to make decisions that could be made by human beings? One word: efficiency. Just as robotic arms increase the speed of assembly lines, computers can increase the productivity of analysts by many orders of magnitude. At Gnowit, a recently-implemented machine-learning algorithm reduced information-processing time from 183 days (when the process was carried out manually) to less than a week. While the cognitive capacity of human beings is impressive, we’re not so good at processing data quickly.
In addition to enabling new features in a variety of online platforms and consumer products, machine learning is currently being used in industries such as retail, healthcare, and insurance. Notice a pattern? Organizations in each of these areas produce large volumes of data. This data contains patterns that can lead to highly-accurate predictions about customer preferences, employee performances, medical diagnoses, appropriate pricing models, and more. The problem is, the data sets in which these patterns are found (usually referred to as “big data”) can be unwieldy, and very large. As a result, they pose a huge challenge for human analysts. This is where computing power and machine-learning algorithms come in handy.
For every organization, there are a whole lot of crucial insights waiting to be discovered. Many of these insights can’t be extracted from big-data. For instance, in finance, subtle signals in online media – both local and international – can be used find investment opportunities and new markets. In marketing and public relations, relevant intelligence from a variety of online sources can help agencies keep track of their brands. Government departments can monitor blogs to spot potential security threats. In every one of these scenarios, machine-learning algorithms can vastly improve the quality of results obtained.
Like the insights buried in huge, complex data sets, those on the internet can be elusive. Web-intelligence tools can locate the online content users need and the relevant information within it; they can also provide analysis and context to turn this information into insight. For a web-intelligence system to be able to do this on a mass scale – which is to say, for such a system to have coverage of a very large number of sources – web crawlers that find and add sources automatically must be implemented. Data scientists can build machine-learning algorithms that enable these web crawlers to differentiate between relevant and irrelevant sources.
This may not sound like a big deal. But ultimately, the result of these algorithms is a much larger scope of results. These results are also of much better quality than they would otherwise be. Everything is improved by the implementation of machine learning – from the accuracy of search results, to all manner of analytics. With intelligence-gathering tools, as with big data insights, machine learning technology holds enormous potential.
Though some media outlets – and Elon Musk – continue to voice concerns about artificial intelligence, many decision makers in the business world have found its benefits too enticing to ignore. Machine learning is a particularly promising area of the AI field. Its ability to greatly increase the accuracy of tools that process large quantities of information is miraculous – especially in a world where we’re drowning in data.
Next time you see a headline decrying the development of advanced technologies, remember the ways in which machine learning is set to make the world a little bit less noisy. And don’t forget who’s in control.
Feature Image Courtesy: Jonty