Every morning, I have an artificial intelligence and machine learning reading list and every morning I see the same 3 problems in AI and machine learning content.
Around 8 am each morning, I make my way to the office, usually listening to the Dan Le Batard show, make my incredibly delicious Bali Blue Moon coffee in my Aeropress and I sit down to look at my aggregated AI and ML articles.
And here we come to problem # 1:
Problem # 1: Bad Headlines!
The first problem I notice is the headlines of some of the best and most interesting articles.
Here’s an example from this morning on Medium
Do you know what Numpy is as a new reader?
How about I/O?
I won’t even mention back propagation.
One of the Big Picture problems with marketing AI and Machine Learning is that the marketing materials are WAY too technical and complicated.
The Waymo article is REALLY close to having a good headline all they need to do is eliminate the term Google I/O re-cap which lowers the value of the content and is confusing if someone is looking for a more accessible easy to understand article which it is.
When it comes to writing headlines you want to take a cue from the high traffic sites like Buzzfeed.
An easy way to start creating more interesting headlines is to use this old Gawker trick: Begin with a question that would false if you removed the question mark.
You can also use Co-schedule’s awesome headline analyzer tool for free Here
Problem # 2: In Article Algorithms and terminology keys.
I come from the Ivy League world of academia, I understand the drilled into you process of showing your work in academic papers.
Guess what blogs and content marketing pieces aren’t
I’ll give you a second.
Did you guess academic papers?
Because academic papers make TERRIBLE content.
You don’t need to show your work in content pieces. All you need to do is show the BENEFITS of your product and those who are interested and so technically inclined will figure out how to get in touch with you and have their data scientists call your data scientists so to speak.
I know it can feel scary to strip out what you believe is the most important parts of your content but it’s not what the audience is trying to hear.
Your ideal clients want you to talk about what your AI or ML can do for them and their business.
Trying to show them HOW it works is like trying to sell hot stew during the Pawnee Summer
They don’t care about how brillant you are.
Problem # 3: There is no “Toaster”
My friend the super smart marketer Tony Almeida has a concept he calls the “Toaster” for his content.
The concept of a toaster is that it is one easy to understand or implement idea that the person reading or watching your content can take away and use without even reading the whole article.
He calls it a toaster after the way that banks used to give out free toasters in order to get new clients to join the bank.
One of the things I’ve noticed from reading about an hour of AI and ML content a day for the past 2 years is that if the article or video does have a “toaster” it is usually in the form of some algorithmic breakdown that you need a CS degree to even begin to understand.
For every piece of content you create you want to have a toaster.
And you need to make that Toaster simple enough for anyone who might stumble upon it to understand like this short video on where AI uses sentiment analysis to determine that Ben Affleck is in fact displaying the emotions of sadness in the infamous “Sad Ben Affleck” video.
Now that is a GREAT toaster that anyone can understand.
Ok so let’s quickly re-cap the 3 big problems I see in even good AI and ML content
Having read a ton of AI and ML content over the last few years I noticed the following 3 big problems:
Bad Headlines. Either Too technical or not focused enough on the benefits.
In article algorithms and terminology keys. Just no.
Not enough “Toasters” or overly technical toasters.
So now you know to spice up your headlines, remove the algorithms and terminology keys and include 1 simple easy to understand idea or take-away in each article.
If you do just those 3 things you will see a huge leap in the quality and reaction to your ai and ml content.
Fear of having jobs taken by robots (or immigrants, or people of color if we’re being honest) is a fear as old as jobs themselves.
A quick search for the root of this fear leads to results like this wikipedia entry for Technophobia which can trace it’s roots all the way back to a group of weavers destroying machines in 1675.
The fear of robots specifically goes at least “The Brazen Android” a story published in The Atlantic all the way back in 1891.
Whether it’s weavers, or Ford assembly line workers or now sportswriters the fear of robots taking our jobs is nothing new.
For the most part the fear of robots was limited to more manual tasks like assembling Ford F150s.
Companies like Braincorp have created automatic floor sweepers, while the Roomba robot vacuum cleaner has been cleaning homes with questionable effectiveness for over a decade.
Robotic Process Automation or RPA use machine learning to complete tedious paperwork tasks such as compiling documents for signatures or checking for compliance or legal privilege.
But for the most part those who created things (like content) believed that we would be safe from the robot job apocalypse.
Only it appears we were wrong.
First the content creating robots started by writing ads. If you’ve worked in online marketing over the last few years you’ve no doubt discovered programmatic Pay per click (PPC).
Programmatic PPC is an artificially intelligent system that allows you to automatically buy, place and optimize display advertising.
Or in other words the robot buys, places and optimizes the ad for you.
There goes a whole industry of PPC consultants.
In fact it’s been estimated that in 5 years all PPC ads will run, and optimize automatically after the initial setup without any human interface necessary at all.
Which is why Google is actively calling everyone who advertises with them to offer them help with their programmatic system.
I know, I’ve been on 3 of those calls and grilled them for information about their robots which I’ll share in an upcoming article.
Now content creators, bloggers, and video makers have another type of AI to content with as content creating algorithms have begun to slowly trickle onto the market.
The technology being used is
Natural language generation (NLG) similar to Natural language processing (NLP) this is the task of generating natural language from machine representation systems.
Intelligent narratives sometimes also called “data driven narratives” which are stories created from the collected data personalized for the audience.
Automated storytelling technology which is responsible for the first AI created screenplay Sunspring
And this technology is ALREADY being used.
3 companies already using A.I to create content:
Quill. The most well known NLG software is Quill which was created by the Narrative Science company. Quill started as an experiment at Northwestern and currently produces over 1 million words a day as it creates reports, news stories and headlines for companies such as Groupon, and T.Row Price.
The Washington Post. It’s no surprise that a paper owned by the World’s Richest Man Jeff Bezos would be on the frontlines of AI created content. The Washington Post has an in house AI content creation system called “Heligraf” which wrote over 850 stories in 2017 alone. Heliograf also composes social media posts, news updates and alerts. Notable events covered by Heliograf include the Rio olympics and local political races in the DMV area.
The Associated Press. While many people associate the AP with the old world journalism of newsrooms, black coffee and unfiltered cigarettes, they are actually one of the most advanced companies when it comes to content creating AI. The AP uses Automated Insights to take care of it’s oft neglected by humans corporate earnings reports beat. But what really separates the AP from the rest of the pack when it comes to intelligent content creation is with Wordsmith “The World’s First Public Natural Language Generation Platform. Let’s explain because that is a mouthful. Wordsmith works with you to create a story. First you set up rules, a template and and the required datapoints. From there Wordsmith does the rest creating the story. Wordsmith generates over 1.5 Billion Pieces of Content A Year or about half as much as I do. Companies that use Wordsmith include Microsoft and Allstate.
If you like me make your living creating content those last few paragraphs sent a cold shiver down your spine.
But there’s still reason for optimism if you are a content creator!
Why Content Creators shouldn’t panic:
Emotional Depth. If you look at the reviews for Sunspring the AI created screenplay you’ll see a lot of words like quirky and interesting. That’s because it’s still very difficult for AI to recognize the role that emotions play in making content interesting! Think of a show like “Parenthood” which is basically emotional porn. They crank the 70s music, a mother and son have a universally emotional moment and we all cry. AI has a LONG way to go until it is capable of writing shows like “Parenthood” or it’s evolutionary successor “This is us.”
Context. In this Post I talk about how the biggest challenge for AI when it comes to language is context. Identifying things like sarcasm, and irony are still difficult for machines to do and are a BIG part of creating content that connects with your audience. Until computer programs can understand something as complicated as the “Aristocrats” joke comedians and other content creators will still be safe
3. Creativity. Last but certainly not least is creativity. Movies like Memento, or Pulp Fiction which bent the way narratives are used can’t be created by AI. AT least not yet. There will always be something to be said for the genius of human creativity that cannot be replicated.
I think about AI with regards to content creation like the famous quote about 500 monkeys at typewriters writing the best novel of all time. Or more accurately like this simpsons clips
While content creation robots are coming, all they are going to do is take the menial content creation jobs. They won’t be writing the World’s best novel or even the most popular blog posts.
If you’re a content creator who just churns out crap, then yes your time is probably coming to an end because the robots can churn out crap faster and cheaper.
But there has always been a market for good writing, entertaining videos and engaging speeches and I think there always will be.
Anyway one of the things that is really interesting about the marketing in the AI world is that IT ALL SOUNDS THE SAME!
This is something I have noticed in very technical or academic types of businesses. These sorts of businesses get very concerned with sounding scholarly or academic and end up sounding out of touch.
One of the big reasons for this is something called “The Curse Of Knowledge.”
The curse of knowledge is a cognitive bias which states that the more you know about a subject the more likely you are to explain it in a way that assumes a high level understanding in the listener.
So for example an AI company might do text analysis and text analysis is kind of like the swiss army knife of Artificial Intelligence it works for a variety of industries from law to human resources which means there are a lot of companies that this AI firm could do business with.
Not only that, but the guy(or girl though it’s rare) who founded the company did his/her thesis paper at some super fancy tech school on text analysis. They have been working on this stuff since he/she was literally a teenager. We’re talking 15-20 years of daily obsessive focus on text analysis. So when he’s thinking about marketing, he REALLY wants to show off all of the high level cutting edge functions of this technology.
And he/she forgets that their audience is not nearly as advanced.
If you look at the AI marketing that is currently being done, the voice is academic, and technical.
It’s writing that is designed for academia not sales.
So, how can AI companies find their voice?
I have a few suggestions:
Don’t let your engineers or founders create marketing materials. One of the most frustrating things about working with AI companies in 2018 is that they have marketing departments that are staffed with engineers and technical writers not marketers. There is a strange distrust and quite frankly and obvious lack of respect for the marketing profession at most AI companies as if somehow marketing is a “soft” science as opposed to the more rigourous computer and data science needed for AI. Right now having technical content is not hurting these companies but it will VERY soon. 85% of executives surveyed by the Harvard Business Review predicted that AI is going to change their companies by 2021. This means that in order to grow AI companies are going to need to stop focusing on Fortune 1000 companies and start focusing on SMBs who are not going to have anyone capable of doing some of the equations in these white papers.
Find the fun. One of the best examples of AI marketing done right is Soundhound which showcases it’s unique Houndify voice enabled platform in everything from cars to break dancing robots at CES in 2018. Instead of focusing on the science behind their platform, Soundhound put the focus on all of the fun and entertaining ways their platform could be used. People gravitate to content that is enjoyable and fun. As Bomani Jones once said ” No one can regulate how much fun you have at work.” AI companies should find the fun both in the office and in their marketing.
Identify and Write To Your Evangelists. One of the interesting things about technology and media is that they need evangelists to spread. I personally learned about Spotify, Uber, S-Town, and Serial from friends and media personalities I trust. When it comes to technology because the learning curve is usually so steep the early focus has to be on finding and creating evangelists both in the press and in the business world. Quick question, right now who is the most trusted writer in artificial intelligence? Chances are you couldn’t think of one because there really isn’t one. Furthermore there isn’t even a group of gurus who you can trust like there is in marketing or other industries. The group of experts are a bunch of data science PHDs who have trouble explaining these concepts without using words that the audience has to look up. Right now when you google AI talks or videos there’s an 89.9% chance you’re going to get a PHD up on stage. And who do academics talk speak to? Other academics. At some point the cycle has to be broken and companies have to identify who the AI evangelists are in the media and who knows how to market AI to SMBs and other companies that didn’t even realize they were the ideal audience.
Pick a niche and produce the best AI content in it. Right now another problem for AI companies is that their products can be used in too many different ways which means that the same company might have a white paper on using AI in sports and a video on using AI in medicine. Right now there is an incredible opportunity to own parts of the AI world. Companies need to focus their content and the voice in their writing and videos on being the best company at a certain niche whether that’s being the best drug discovery AI firm or being the AI firm that handles legal briefs. Once you define this niche it helps you to define what your content is actually about and frame things appropriately.
I’ve tried really hard to avoid saying HIRE CONTENT PROFESSIONALS because I understand that right now the online sales market for AI is not really there and all sales are made at the C level face to face or through long arduous government contracts. But the point stands. Before these companies know it, the time to sell to SMBs and focus on widespread adoption will be here and the companies that take their time in finding their voice, dominating a niche, and continue to let academics create marketing materials the further behind they will fall.
Happy Tuesday, that’s a set of words you don’t hear too often.
I am enjoying the beautiful 91 degree weather out here in Las Vegas while trying to decide what I want to eat for lunch.
I spend a lot of time thinking about what I want to eat, but unfortunately I am not able to become one of those decision fatigue avoiding robots like Alabama Football Coach Nick Saban who eat the exact same thing everyday.
Just in case you were wondering, Nick Saban (who I call my illegitimate father) eats iceberg lettuce, tomatoes and turkey with I think ranch dressing for lunch each and every day. And he’s rich! Imagine choosing to eat that when you make 9 million dollars a year. But I digress…
In other news, I’m thinking of doing a whole series of beginner’s type posts on A.I and Machine Learning.
I think a lot of the more advanced concepts go over people’s heads and they want to be able to understand and define these terms easier.
The beginner’s series would be aimed at businesspeople who understand that AI and ML are important and that they are going to affect the way they do business in the near future but don’t really have a firm understanding of AI or ML.
These are the executives and small business owners who KNOW they should have a better knowledge of AI and ML but either don’t have the time, or are too embarrassed to admit that they don’t understand terms like Predictive Analysis.
If you know AI and ML are important but would freeze if I asked you to define either term then these blog posts are for you!
Or so I think.
Let me know what you think in the comments below.
And while you’re here be sure to take a look at my latest blog over on medium:
This is the most recent thing I wrote on AI and goes into the process of deep learning using Artificial Neural Networks (ANN). This piece is pretty good but I could have included other examples of deep learning besides the 4 I put in there.
This post was good for beginners in AI and business development. I think looking back on the piece I could have positioned it more to be specific to small and medium sized businesses as larger businesses have been doing this stuff.
This post would be perfectly at home here on the Content Marketing King blog. If anything this posts shows how much data is being collected and how data focused the next generation of content creators will have to be to survive.
I used to write a lot about the job search process because I find the process to be inherently unfair. Employers have all the power and often lie in job applications about the job or compensation. This post accurately lays out how the age of intelligent machines is going to affect finding a job over the next 20 years or so.
This is a pretty basic list of applications of machine learning. It essentially lays out the areas that I would be writing about in later pieces. It is interesting to see how some of these areas like autonomous driving have been in the news lately.
In the first piece I wrote around AI and ML I looked at the idea of predictive analysis and how AI and ML are being used to make predictions about everything from court cases to earthquakes with various levels of success. At this point I was still figuring out the differences between ML and AI.
Ok that’s most of what I have written about AI and ML that is on Medium I’ll have to track down some of the links in INC and Fast Company as they tend to change fast.
I’ll be sure to post more about AI and ML in the future as I continue to find a place for my writing on the subject.
This week I have interviewed the founders or CEOs of 4 different AI and Machine Learning companies for upcoming articles on Forbes and INC.com.
In this interviews I learned a lot of interesting things that might not necessarily fit in the pieces I am writing but I figured I would share here as a sort of “Behind The Scenes” look .
And because I love list posts and I know you guys love to read them I put together my list of
7 Random Thoughts on AI and Machine Learning:
What proprietary data do sports teams have that they aren’t telling us about? We already know a fair bit about Machine learning with sports data but everyone I talked to who worked in sports analytics confirmed that the teams that are using Machine Learning have WAY more data than we have access to on the outside. And I want to know what they have.
How are we going to eliminate bias from legal data when the legal system itself is bias? The legal system in America is biased. One need only google the incarceration rate for African Americans vs White Americans to see that. But there’s also documentaries like 13th By Ava Duaverney which explain it way better than I ever could. To quote the great philospher Kanye West “Face it Jerome get more time than Brandon.” So another challenge arises in the legal field where machine learning data will need to be scrubbed of bias before it can give sentencing recommendations or offer rulings.
What can’t they put a brain in? After talking to the folks over at BrainOS I am convinced that a big trend is going to be installing autonomous brains into everyday items essentially transforming any task that manual from floor cleaning to window washing into a task a robot can do.
AI is going to revolutionize the drug discovery pipeline. The way drugs are discovered and brought to market now is horribly inefficient. It’s almost as inefficient as the Cleveland Browns search for a quarterback (AY-YO). It takes on average around 10 years and a billion dollars of investment to bring a drug to market for a rare disease. Companies like Recursion Pharma which are mapping both drug compounds and human biology at a rate of 7 terrabytes a week are going to completely change how fast drugs are brought to market and disrupt the entire drug discovery pipeline by providing a better service faster.
Why couldn’t the 76ers fix Markelle Fultz’s shot with biometrics? This is a question that kept coming to my mind as I spoke with sports analytics people. The technology exists to bio-map Markelle Fultz’s broken shot but no one I talked to seemed to have an answer as to why the 76ers hadn’t used that technology or why it hadn’t worked.
6. Open Platforms are where the big breakthroughs are going to come from. Similarly to how smaller companies tend to innovate more effective content marketing strategies the companies that are creating open source platforms like Soundhound with Houndify and BrainOS with their brains are going to be poised to let “hackers” and DIYers make the major breakthroughs and then come in early and either copy or buy up those innovations to take to market.
7. Humans have ALWAYS feared Robots becoming intelligent and taking over. The first mention I could find of robots attacking humans goes back to the early 19th century and that doesn’t count the idea of “bronze” or “gold” men who weren’t quite human or friendly attacking all the way back to 600 BC. To me it seems like a guilty conscience. After all if we really believed we were living the right way wouldn’t we think smarter robots would agree? What does it say about us that we believe the first thing the robots will do once they attain sentience is get rid of us?
Next week I plan on writing more articles on machine learning and artificial intelligence including one about how machine learning and artificial intelligence are working together.
On this blog I am going to write about creating content that sells products or services and the differences between the kinds of content that sells a product like a black mini fridge versus the type of content that sells high priced coaching or consulting services.
On Wednesday I’m going to write about what I learned from a year of playing high level poker online and how that relates to business and writing.
And a week from today I will have a post about my how my diet is going which doesn’t really have anything to do with business but I’m down 11 lbs this month and want to brag and talk about what I’ve learned about dieting.
So lots of content to come as always because I remain the king.