SaaS Content Writer & Strategist 🚀 Remote Work & Mental Health Advocate
Artificial intelligence will inevitably be all around us one day. At work, at home, on the street, maybe even in our dreams.
Heck, it’s already everywhere.
So should marketers be preoccupied with learning all about it?
To a certain extent, definitely!
“But AI is so, you know, for techies.”
🙄 No. You’re already using it actively to begin with. So I’m sure you have some kind of understanding of how it works. You just don’t know it yet.
Fancy buzzword or not, it’s definitely not something that should intimidate you.
I don’t even need to list a bunch of benefits. Just keep this in mind: Google’s algorithm is AI itself.
One central use of AI ( data mining and machine learning in this case) is for creating predictions and recommendations for your clients (done using the nearest neighbor classifier if you want a fancier term).
This type of system is commonly used to select songs, videos, movies, items, or other types of content that’s already similar to what one user is commonly searching for. The principle behind this system is that a user is more likely to click on items that are similar to previous searches, thus matching his/her interests.
And it works.
Let’s have a look at some recommendation systems that use data mining and machine learning to help users find the best content and resources:
First up, YouTube. If you’re an avid user of this website you might be able to understand some of the basics behind how their recommendations algorithm works just by looking at what YouTube tells you to watch next. It’s usually videos from channels you commonly watch, videos with/about the same topics or people, or continuations from other videos.
The only problem is that platforms such as YouTube are still testing their algorithms. This is why you’ll often search for something just once or watch one single video on a topic and still get more recommendations for similar videos.
Why? The algorithms are learning. It’s only been a few years since they’ve been learning.
And any algorithm or machine is just like a human. It takes more than 15 years to get a better understanding of the world.
Next, Spotify. You search for certain songs, listen to various songs or podcasts, and replay some more often than the others. In the background, everything is recorded as data that helps the algorithm pick other songs for you and even create complete playlists you might like:
I’ve given Spotify as an example to showcase how for certain fields of interest it’s much harder to come up with accurate suggestions. You can work around genres, artists, and podcasts topics, but that’s about it. You can’t really nail the taste of the listener yet. Just check out your own playlist recommendations and see how many songs you actually like from there. If any.
For a more straightforward example, see Medium which simply sorts out all of the content on their website according to your topic searches and past readings:
A similar example is Zest where content that’s first assessed through machine learning is closely moderated by a human to ensure accuracy (you can read more on the process here):
NOTE: As we’ve seen in YouTube’s case (let’s not even mention Spotify), AI by itself is obviously still not independent. In fact, it’s quite far from being so. That’s why constant human supervision and intervention is a must.
Other platforms with smart, AI-based recommendation systems include Netflix; HBO; Amazon; Pocket; SoundCloud; Shazam; Instagram, LinkedIn, Pinterest to a certain extent; and many more. Even Waze. Ever noticed how after using it regularly the app seems to know when and where you go to work? Let’s not even go into all the details that lie behind optimal route calculation.
Simply being an Internet user is enough to be “concerned” about artificial intelligence and how it’s affecting your life.
In the wrong hands, AI-influenced search algorithms might have many negative implications since they can be fully controlled by companies.
This means they’ll be able to show you only the content (articles, images, news, products, services, etc.) THEY want. In other words, they spread filter bubbles around the web.
We’re so busy looking at how our website or app is performing, that we forget there’s so much data out there for us to use to make predictions and anticipate future customer behaviors and patterns.
Think of all the possibilities that artificial intelligence can bring:
All this even if you’re new to the industry and have just a few months’ worth of data to start with.
Google already added (well, still testing actually ) a series of Analytics Intelligence features. This tool literally learns more as you use it as you can teach it yourself by rating whether the answer you got was good or not.
What the tool will ultimately do is help you find any kind of information you need in your Google Analytics account by simply asking a question.
Here are a couple of examples:
All this data will easily let you know if there’s any kind of problem with your marketing and sales campaigns or where there’s still room for improvement without all the fuss of going through endless views and modules.
If you want to get (or get your devs) into the technicalities, the googleAnalyticsR website will provide an R library that can be used with Google Analytics data.
As a side note, keep in mind that anything you export from Google Analytics is ultimately still a CSV file. That means that any data you have from whatever tool you’re currently using to analyze your stats will work just as well in R or Weka.
Cool. What can it do?
Create buyer segments based on real data not just your predictions, better understand your clients and prospects, generate new leads, create statistics, establish budget estimations. Even mine user sentiment from tweets. You know. So you’re not reading each and every one of the 1000+ brand or product mentions you got.
From the very basic descriptive statistics that tell you WHO your buyers are and what they want, the E-commerce industry has already been using R for years now. Data mining helps online stores predict product sales and cross-sell and upsell products based on a buyer’s previous purchase.
Now for some bomb news: Google’s Brain is also AI. 🤯
TensorFlow-based to be exact. The Google team developed this machine learning system that currently lies at the base of most of the things you see in your search engine, including the ranking algorithm, photo search, YouTube recommendations, Google Lens (object/image recognition), and Google Assistant (voice recognition). It’s really a “magical solution” to any discovery, predictions, perception, or creation problems.
All of these tools are based on artificial intelligence methods such as machine learning, deep learning, and NLP. Think of them as a baby learning its first words and gradually being able to put sentences together, create complete speeches, deconstruct a phrase, correlate one topic or idea to another (*cough* keyword clusters *cough*), and correct your mistakes. [Oh, and knowing the basics of NLP will help you better understand Google’s algorithm changes😊]
It’s also the reason why you can talk to Siri or Cortana.
“Robots” using the language of the people are anywhere from writing tools, social media monitoring software, and CRMs, to the chatbots we’re starting to prefer over humans because they’re always available.
Step-by-step, natural language processing is easing (if not ending) our work. Not just in marketing. Customer support, recruitment, sales, writing. All these industries are starting to see its impact.
From your friendly Messenger bot that takes over a conversation until a human is able to respond:
To a headline analyzer:
IBM Watson’s Tone Analyzer:
Even smart reply suggestions you get on Gmail or LinkedIn:
And “simple” tools that summarize text:
Before you sign off, don’t forget to share your thoughts on the topic and remember, you’re being watched. 🤖