Amazon believes that the future of machine learning — the idea that software can be programmed in such a way that it teaches itself and eventually becomes artificially intelligent — is in shopping.
That seems to be the hidden message tucked inside the launch of its new Fire Phone.
Machine learning is the hot new research area in tech. The largest class at Stanford University right now is in machine learning. Most people envision machine learning as being about intelligent user-interface systems, like the operating system with a personality in the movie "Her." Or robot soldiers that can plot tactics and strategy for themselves. It's one of Larry Page's top priorities for Google this year.
But Amazon CEO Jeff Bezos thinks machine learning might be for shopping.
The phone has two dedicated shopping/consumption functions, all designed to funnel more mobile commerce through Amazon, of course. There is the Amazon Prime membership, which gives users free movies.
But of more interest is Firefly, Amazon's real-world shopping recognition technology. Customers can use the phone's cameras to scan physical objects in the real world — like a store — and Firefly will identify them, and then provide a channel through which users may be able to buy the objects more cheaply.
Baird Equity Research analysts Colin Sebastian and Benjamin C. Gaither noticed that Firefly relies on machine learning to figure out what nonstandard physical object you're pointing your phone at, regardless of the location and lighting, and what manifestation of the product it will show you on the phone as a reference. They wrote in a note to investors:
Using machine learning to create a mobile competitive advantage. Through the incorporation of the newly announced Firefly technology, Amazon is attempting to reduce some of the friction in mobile commerce by enabling users to scan phone numbers and email addresses, recognize music and video content, and identify over 70 million physical products with the push of a button. Given Amazon's wide competitive moat in ecommerce, this enhanced mobile functionality could help leverage an already disruptive retail platform, and it's a feature unique to Fire Phone.
Full of machine learning. Key distinguishing features of the Amazon phone include the Firefly button, which acts as a visual search engine (image and sound recognition), and doubles as a product scanner leveraging Amazon’s database of over 100 million items. We believe the most practical application of Firefly is shopping for physical or digital items, using Firefly for identification, information and comparisons. Some of the Firefly functionality is already available in Amazon mobile apps (for example, product scanning), and presumably Amazon will eventually integrate Firefly with Fire tablets.
Do not underestimate the resources Bezos is throwing into machine learning for shopping. Amazon is virtually permanently recruiting machine-learning scientists (you can see the vacancies advertised here) and those jobs start in the six figures.
And it's not just Amazon, where Bezos expects machine learning investment to occur. The Fire API and software-development kit will allow app developers to plug their own machine learning into the system if they want to use that instead of Amazon's native version.
You'll notice that the ML involved in Firefly is essentially a souped-up version of a search algorithm. Instead of punching in keywords to Google, the camera scans what it sees and then displays its best guesses for results. This is a version of so-called semantic search in which an intelligent machine attempts to give you the results it thinks you want rather than a mere fixed response to a fixed query. Some people think Google has a long-term weakness in search, because it is trapped in the currently lucrative business of responding to text keywords rather than delivering meaningful "semantic" results before we know exactly what we're looking for.
There are lots of little niches in search where Google is weak. App search, for instance. Or music recognition. Both Facebook and Google are working on competing machine learning products.
Meanwhile, Amazon and Google compete head to head in shopping searches, which is one reason Amazon wants another platform on which Google's ads don't appear.
This time, the platform will be intelligent, it seems.
SEE ALSO: This Google search for 'coconut flour' shows why Amazon needed to make a smartphone
SEE ALSO: Facebook is building a massive new business that exploits a key weakness at Apple and Google