Defining Artificial Intelligence and Machine Learning

by JP Sherman

First, I want to describe the subtle differences between artificial intelligence and machine learning.

Artificial intelligence is a high level concept that describes the ability of machines to do two primary things: independently perform tasks and make decisions based on the data it has and to synthesize data to make a unique decision based on that input.

Driverless cars are a very practical application of AI in action in the real world. A more search-focused action of AI is Google’s RankBrain processing signals from a site or page to create a unique ranking factor set that can satisfy the intent of the user.

Machine learning is the application of an AI rule-set to allow machines to consume data and learn for themselves. A good example of this is how Google understands entities. For example, Google understands that Batman is an entity. This entity is associated with locations, enemies and allies.

Once the machine knows that Batman is associated with these categorizations, it is programmed to find entries that fit those categories. The machine searches for “locations” and comes up with “Batcave”, “Gotham City” and “Arkham Asylum”. Its search for enemies comes up with “Joker”, “Two-Face” and “Bane”.

The machine is building a template to understand the relationships between Batman and other entities. As the machine starts to collect data on associations and relationships, it can use the template it built for Batman to start filling in data on other entities.

The critical thing to understand is that AI is the set of rules that perform intelligent tasks independently and machine learning is the application of those rules.

So, how do we, as search professionals, apply artificial intelligence and machine learning to influence ranking and deliver what users intend to the SERP?

The most accessible way for us to influence results is to use structured markup, powered by a robust taxonomy. The good news is that most CMSs have a way to natively build a taxonomy. Think of taxonomies as a way to define a “thing”: a product type, a content type or any other high level category.

Building a taxonomy will create a unique definition that lists the unique attributes of what you’re defining. I highly recommend having a working knowledge of the Dublin Core Metadata Initiative.

Once you’ve built a good taxonomy that defines things, use that data to build structured markup schema, using attributes from Schema.org.

I know that I’m one of those people who bang on the “structured markup” drum but the advances in AI and ML will primarily affect the organizations that are structuring and defining the data they have in the format that the machines can consume easily.

This is the point where I step away from search engine optimization and point to digital assistants like the Amazon Echo, IBM’s Watson and a host of other non-search engines that consume web data to understand entities and relationships in order to synthesize information.

Structuring your data cannot be viewed with just the perspective that it’ll give you a greater position on Google’s SERP, it needs to be viewed with the idea that, if done well, your data can reach and influence a greater audience than ever as machine learning expands into your home, your car and many other invisible areas of your life.

JP Sherman is a ten+ year veteran of Search, Findability and Competitive Intelligence. JP works as the Search & Findability Manager for the Red Hat Customer Portal. His responsibility is to bridge the intention gap between hundreds of thousands of technical and support documents and the customers looking for them in Google and Red Hat’s internal search platform.

How Virtual Assistants Can Effect Rankings

John Leo Weber is the COO at Geek Powered Studios, a full service digital marketing agency in Austin, TX. John helps to craft the digital strategy for all Geek Powered clients and writes about marketing for a variety of publications across the web.