Going deep with deep learning: Martech insights, action & impact
Columnist Andy Betts walks you through the ins and outs of deep learning and explains why it's set to be a game-changer for marketing programs.
Andy Betts on January 2, 2018 at 10:47 am
artificial intelligence and machine learning is picking up pace, and those who thought adoption was a few more years away are finding the reality much different. AI is already embedded in our everyday lives, and forward-thinking marketers are embracing machine learning technology efficiency to automate and scale their marketing programs. However, little has been mentioned of deep learning, the gem in the AI armory that provides even more powerful insights to marketers. In my last article, I focused on AI applications and explained why not showing up in 2018 without an AI system or application in your martech stack could leave CMOs lagging. Taking that premise a step further, not utilizing deep learning technology means that marketers are losing out on essential insights that I predict will fuel marketing technology development in 2018, 2019 and beyond. Did you note how I used the terminology “application” with AI and machine learning and “technology” with deep learning? The reason for that is to make a key distinction, as deep learning is a combination of big data sets, machine learning, computer processing power and neural networks that make applications smarter as it learns. It’s important that technology marketers distinguish between AI and the subcategories that make up AI. Artificial intelligence (the umbrella term) is the science of making machines smarter, which, in turn, augments human knowledge and capabilities. Artificial intelligence is any computer program that does something smart. AI is normally categorized into narrow AI and general AI (where machines can do exactly what humans do — think robots and the film “Ex Machina”). Today, only narrow AI applications are available, and these can be either supervised or unsupervised as applications grow in intelligence. Machine learning is a subset of AI where machines take data and begin to learn for themselves. Large data sets feed algorithms that are programmed to learn and improve without the need for human data input and reprogramming. Machine learning allows a system to learn to recognize patterns on its own and make predictions. Deep learning is another subset of AI that is better described as a technique. Deep learning technologies train themselves and are based on the biology of the human brain and neural networks. Massive data sets are combined with pattern-recognition capabilities to automatically make decisions, find patterns and power self-learning.The growth of
Going deeper with deep learningDeep learning is the gem in the AI armory — it is the brains behind it — and has enabled many practical applications of machine learning and powered the growth of AI. Deep learning is a technology that makes applications smarter and more natural as it makes sense of big data via immense computer processing speed. One of the biggest differences between machine learning and deep learning is the number of data sets used and data points involved. Machine learning uses thousands of data points, whereas deep learning uses millions. Think about that for a second. For a marketer, the potential output and insights from deep learning are potentially staggering. In the marketing technology space, where competition is fierce and marketers are always looking for an edge, this may be just it. Deep learning originally was associated with self-driving cars, surveillance and science fiction, but it’s now being associated with AI, machine learning and major marketing technology players. Deep learning is basically an advanced type of machine learning.
The evolution of deep learningBack in 2014, Google bought AI startup DeepMind for more than $500 million, and in 2016 Microsoft set up the Microsoft Ventures VC fund, which focuses on investing in AI companies. Microsoft also joined with other tech giants such as Apple, Amazon, Google, Facebook and IBM to work on the Partnership on AI, a consortium aimed at conducting research and establishing best practices. Between $26 billion and $39 billion was spent on AI in 2016, according to McKinsey, the bulk of which included investments in R&D and deployment by the likes of Baidu, Amazon and Google. That same year, Uber signaled its deep learning intent by acquiring Geometric Intelligence, an AI startup working to explore beyond the boundaries of what machine learning has to offer. Meanwhile, Facebook is creating deep learning AI which aims to find out what matters most to Facebook users, while Salesforce.com is working with MetaMind, an AI startup it acquired that specializes in deep learning. IBM has Watson, and Adobe has Sensei, both of which focus on machine and deep learning. Machine and deep learning intelligence is being built all around us. (See image below.)
Deep learning and marketing technology applicationsThe biggest challenge that marketers face today is the overabundance of data.
- There is simply too much data out there to process, segment and sieve learnings from. At least 99 percent (PDF) of big data is not analyzed, according to the IDC. Take a moment to think about that if you can.
- Data Management Platforms (DMPs) try to tackle this problem by providing insights from large volumes of data, but the reality is that many tasks still need to be performed (and supervised) by humans.