(By Richi Jennings)
“But here’s the rub: No one, not even the people who built the model, entirely knows how it works. … Algorithms that “mine” the data for patterns are already being used in every stage of business. … If mathematical models make useful predictions, does it matter if nobody understands them?“
Suddenly, everyone seems to be talking about deep learning. Or at least, that’s how it seems to me.
What is it? It’s a powerful ‘new’ way to analyze big data that allows computers to learn, without being taught. As with any new buzzword, there are plenty of chattering commentators who don’t really understand it.
So, I’ve got some explaining to do…
On the one hand, it’s still early days: This is no overnight revolution. On The Other Hand, the Sci-Fi dreams of artificial intelligence seem much closer they did last week, thanks to the tireless work of people like British researcher Geoffrey Hinton (pictured).
Deep whatnow? Alistair Croll is first to try to explain:
[It’s a] hot topic in big data. … Teaching machines to think. … Rather than teaching a machine explicitly…Deep Learning uses simpler, core ideas and then builds upon them — much as a baby learns sounds, then words, then sentences.
They’ll likely supplant the expert systems…used in many industries, but have fundamental flaws. Ben Goldacre [said] almost every patient who displays the symptoms of a rare disease instead has two…common diseases with those symptoms. …this is why House is a terrible doctor’s show.
And Brian Matchick also has a go:
Computers are great at a lot of things, but they can’t grasp the basic tasks routinely performed by a 2-year-old child. [But] “deep learning” is showing a great deal of promise…making self-driving cars and robotic butlers a real possibility. … They are still limited, but what they can do was unthinkable just a few years ago, and [it’s] advancing at an unprecedented pace.
The computer is given a huge pile of data and asked to sort the information into categories on its own, with no specific instruction. …the machine can grasp the low-level stuff before the high-level stuff. …but we still have a long way to go. … Deep learning has no way of establishing causal relationships or making logical inferences.
Self-driving cars? Ugh. Aren’t there any better examples of business applications? Richard Waters runs deep:
New tools used in scientific research have accelerated the process of discovery.
The ability to analyse massive data sets and use “deep learning” in computer systems that can adapt to experience, rather than depending on a human programmer, have led to breakthroughs. These range from drug discovery to the development of new materials to robots with a greater awareness of the world around them.
That’s more like it. Several commentators point to Apple’s Siri as an example of deep learning, butParmy Olson says that’s not so:
Google, by contrast, is moving toward so-called deep learning technology pioneered by…Geoffrey Hinton, whom Google hired in early 2013. [It] promises superior results.
The company is meanwhile moving towards greater personalization of its search results.
You can say that again. David Amerland explains:
Search, as we know it, is dead. [But] semantic search enables a new phase of the Web, where we stop searching and start finding…information directly related in context, not just in keywords.
[It] marks the shift from search results based on probability to results based on actions and an understanding of natural language.
Do you want your products and services to be found online? …the strongest currency is great content. It buys you eyeballs and minds.
And Emma Byrne offers another Googley example:
Google’s Flu Trends project is a prime example of data mining. Google has billions of bytes of data about its users. [It] used [deep] learning to learn to predict flu epidemics [finding] a set of less than one hundred variables that would predict where flu would strike next.
But here’s the rub: No one, not even the people who built the model, entirely knows how it works. … Algorithms that “mine” the data for patterns are already being used in every stage of business. … If mathematical models make useful predictions, does it matter if nobody understands them?
In tomorrow’s business, big data can tell you more about your operations than your people alone.
So how can you take advantage of these ideas? Lee Congdon has this broad-brush advice:
In a way, almost every company is becoming a technology company. … Your IT team can help figure out what your customer base wants, why they want it, and what challenges they face.
Use technology to expand your business horizons. Attract a new generation of customers or dazzle the existing ones.
OK, good, but more specifically, please: How does a business IT department jump on the Deep-Learning train? Derrick Harris has one idea:
Google silently did something revolutionary on Thursday. It open sourced a tool calledword2vec, prepackaged deep-learning software.
Geoffrey Hinton…pioneered the use of deep learning for image recognition and is now a distinguished engineer at Google. … Google calls it “an efficient implementation…for computing vector representations of words,” [using] two new natural-language processing techniques. …it can recognize the similarities among words (e.g., the countries in Europe) as well as how they’re related to other words (e.g., countries and capitals).
If this is all too esoteric, think about these methods applied to auto-correct or word suggestions. … Using deep-learning-based approaches, a texting app could take into account the entire sentence, [having] a better understanding of what the all words really mean in context.
So, powerful new ideas for learning new things from big data. But as I often find myself writing here: Step One is to store it!
“Opinion pieces of this sort published on RISE Networks are those of the original authors and do not in anyway represent the thoughts, beliefs and ideas of RISE Networks.”