It’s the Machine Learning Products, Stupid
Products: What Actually Moves the Needle with Machine Learning
The ascendence of big data (already an antiquated term) initially has resembled a gold rush. It was if everyone suddenly became aware that data, especially consumer behavioral data, was extremely valuable. Consequently there was a stampede towards data collection and retention. However, the capacity to actually leverage that data was initially unexplored.
Then machine learning started to emerge as a natural application of that data. Now it seems like a million startups have blossomed that use machine learning as their secret sauce. There is substantial hype.
With machine learning, as with all new technology, what really matters is the products. It is these products that will actually move the needle, transforming society in ways that can be difficult to predict. When it comes to what really matters in this machine learning narrative, it’s the products, stupid.
Consider Amazon’s Alexa platform. Alexa isn’t the most impressive example of machine learning technology: you have to address it in very specific ways that come wth a learning curve. I often tell people to consider their Amazon Echo as a smart Collie rather than a human. (Lassie! Re-order toilet paper! Go girl!)
But Amazon got certain things really right about the Alexa platform. A voice activated, in-home unit, attached to Amazon’s e-commerce stack, that was expandable through third party skills right from the start, and was accessible through relatively cheap hardware. From that foundation, Amazon can expand into tons of different applications. The Alexa platform is a great product embodiment of machine learning technology.
Unfortunately there’s still not enough cross-communication between machine learning experts and product managers. Much of the applications for machine learning algorithms remain opaque to your average product manager (how is a linear regression going to help me find movies I’ll like?). Product managers need to learn more about the practical applications of the technology in order to shepherd the next generation of machine learning products.
We probably need some sort of Machine Learning For The Non-Technical book or resource. A conceptual framework that looks at machine learning algorithms as building blocks for applications, so product managers can think in terms of Algorithm A + Data B= Feature C. There are many other earlier examples of technological shifts making their way to product road maps. For example, the camera in the first generation iPhone wasn’t good enough to support barcode reading applications. It was only once the hardware improved that a thousand product ideas based on barcode scanning were released.
Of course, the quality of a physical component such as a camera is relatively easy to understand compared to machine learning, which at its heart is a cluster of algorithms applied to data. The abstract nature of it creates a mental barrier for the non-technical, rendering it into a kind of magical black box. Unfortunately, this opacity slows the advance of products that leverage machine learning.
So how does this situation progress? What usually moves things along is people learning from others’ examples. For example, the rise of AJAX in web applications, was jump-started by Google Maps. As more examples of applied machine learning emerge in products, we’ll probably see a virtuous cycle develop where product managers will learn to leverage the technology for themselves. The resulting products will show the real impact of machine learning on the world.
I’m Loren Davie, CEO of int18 and a 20-year survivor of the tech industry. I’m teaching a Product Management course at Byte Academy this summer. You should check it out.
