This is Why Deep Learning is Your Future
by BRADY EVAN WALKER
Deep Learning is More Than a Buzzword
Deep learning powers the next generation of uncanny tech: image recognition auto-tagging you in places you shouldn’t be. AI assistants that order pizza, beatbox, and tell politically correct dad jokes. Badasses kicking butt at chess, Jeopardy, and Go.
But what is it?
Deep Learning is the Next Step in Machine Learning
Machine learning is based on the idea that a single algorithm can give insight into a dataset without needing to code for every problem. In other words, the algorithm creates and builds upon its own logic. An everyday example is a spam detector that “learns” common features of junk mail.
This ability is called feature extraction. The feature could be simple, like color or musical tones. It could be much more complex, like facial recognition.
For marketers, one of the most promising deep learning technologies available is Ditto Labs. Ditto offers a detection system to identify brands and logos in social media photos. Their software determines salient features of the environment in which brands appear. It even analyzes the sentiment of the people photographed.
To explain deep learning, let’s first talk about Artificial Neural Networks, a computer software model based loosely on brain structure. Each node or “neuron” is a statistical learning model that estimates functions or features.
In an Artificial Neural Network, each node tackles simple functions. The nodes are organized into a web and feed into each other to increase processing efficiency and to tackle high volumes of input. In a deep learning model, each node is assigned a very small feature extraction task. The combined effort results in a more comprehensive computer education.
Instead of hand-coding commands for holding a conversation, for instance, Google teaches its AI assistant with reams of screenplay dialogue from which to extract, interpret, and learn.
The Future of Machine Learning
There are tons of things we’d like computers to understand that cannot be boiled down to objective functions. But the current uses are increasingly astonishing.
Affectiva, a deep-learning powered image analysis technology, can report back on the emotions on a person’s face in any given photo, which is used primarily for market research. It has other applications, though. Hershey is currently using their technology in store for “Smile Sampler” promotion, a vending machine that doles out confectionary only for sincere smiles. That’s right. The program knows if you’re faking.
Such technology could, one day, power brick and mortars in more integrated ways.
Would a computerized image analysis program spot you as you walk in the door of your favorite store? Then boom, your profile with purchase history pops up on a customer service tablet?
According to an analysis of your social media profile, the sales clerk knows you wear primarily black, knows your size, knows you never wear sandals, and that you generally wear a bolo but forgot to put it on this morning. This may sound creepy but with the right mix of technology and customer-centric attitude, great companies will make this feel like a differentiated, personalized service.
Boston-based Indico offers deep-learning text and image analysis tools. Their text analysis works as you type to analyze sentiment, politically charged language, and other topics and keywords. It can even predict Twitter engagement.
This is a powerful tool for PR and copywriting. Writers of all stripes (like me) already use programs like Grammarly and Hemingway to supplement their instincts. Indico and similar technologies are the next step in computerized mentoring for writers.
Marketers also have a fresh opportunity to join forces with deep learning platforms, just as Domino’s Pizza was the first company to integrate with Amazon Echo smart home assistant. But it’s not just a matter of being buddies. As TechCrunch reports:
The feature is actually being made available by way of a third-party application from delivery chain Domino’s, which has been experimenting with new digital ordering systems over the past year, including the option to order pizzas by tweeting an emoji.
So even if you don’t have the budget for your own deep learning investments, you can push your innovation budget to expand customer service and accessibility in the digital realm. Otherwise, you’ll be left behind. The digital revolution is still turning.
While experts disagree on the timeline, few seem to doubt that deep learning technology will one day reach a level of sophistication that will change the world, powering everything from hyper-intelligent personal assistants to self-driving cars, inhabiting the offices of knowledge workers and running entire factories and warehouses. It’s already changing the face of marketing.