The term artificial intelligence is gaining in popularity and is already keeping many people up at night. Will I wake up tomorrow in a normal world—or will robots have taken over my job?

The forecast for tomorrow looks fairly stable. Most articles about artificial intelligence are essentially media hype—tasty, charming, readable, but overly inflated.

In the case of the Morgan trailer, it was in fact a fairly standard computer program that, with the help of experienced experts and using deep learning techniques, simply analyzed 100 horror trailers and stored the patterns. Then it selected six minutes of footage from the film that best matched those patterns.

Is pattern searching and comparison considered artificial intelligence? Most amateur statistics enthusiasts would more likely classify it as data mining using highly advanced tools. For example, linear regression.

Among the experts in this field, such techniques are classified as supervised deep learning. A specialist (with lots of coffee and willpower) trains the programmatic solution to reach a conclusion.

Artificial intelligence, in contrast to supervised, would mean unsupervised deep learning, where the program trains itself. And primarily for its own sake, that is, for its own needs. So far, no one has even come close to this.

If the terminology has motivated you, here is a guide for your career as co-creators of deep learning.

However we refer to this concept, it is incredibly effective. Watson performed the movie analysis and selected the best clips in 24 hours.

For an experienced human creator, the same work would take about 15 to 20 working days. Watson can now repeat this anytime in just a few hours if we provide him with a horror movie.

And Watson is not the only one at the table who would eagerly grind your coffee incredibly fast:

• Google Analytics analyzes the behavior of millions of users and stores behavioral patterns. PaveAI autonomously converts this data into decision-making recommendations
  
• Facebook Messenger answers questions from your users.
• AdWords suggests the most suitable images, ad layouts, and their appearance based on the content of your landing page.
  
  

All of these activities just a few years ago required entire marketing teams.

From trailer to full movie

Programs for film creation will soon independently generate trailers using deep learning. Human interference will probably only worsen the result. You know what happens when your boss last meddled with the design of a PowerPoint presentation?

And soon they will intervene in the production of recorded scenes.

Computer programs can create new footage from existing shots — changing the text, pronunciation, emotions.

In just a few steps, we can foresee that a program will read a book and, based on it, create a movie. Whether a thriller or a romance. It could even let you choose your favorite version of the same film, say, a comedy.

Inefficient people in marketing

In this near future, will there still be a need for people behind the cameras and microphones, stylists, set designers, and all the others who currently make a living from creating films or video ads?

The economically obvious answer: a few kilowatts of electricity is simply a much lower cost than an entire team of experts who constantly need coffee. And more versions of a film mean satisfying a broader group of users.

The same applies to marketing:

• Would you pay a designer if a program can create the optimal ad in a few minutes?

• Would you pay a copywriter if a program can write an exceptional text in a few minutes?
• Would you pay a marketing manager if the program can devise a better marketing strategy?

The cost of deep learning AI solutions

It’s quite clear that the huge human cost comes from coffee.

 But in deep learning systems, there’s a similar dependency: input data. The program can generate some on its own, but in the real world, the combinations are endless and the parameters keep changing.

We need fresh (as accurate as possible) data in large quantities. Coffee, anyone?

As a result, Google needs real users for its deep learning solutions, it has to acquire and retain them. It has to ask for participation, look for legal loopholes in storing and processing that data.

Sound familiar?

Exactly, marketing isn’t free, and data acquisition is a cost for Google or any other deep learning system.

The data market behaves like any other market. With demand, the price rises. It depends on how many users are willing to share their real data.

Scope, data relevance, and accuracy

A deep learning system gives quite accurate solutions when we have a large volume of data — or very wrong solutions.

But what if you’re a small café wanting to advertise your offer in a village? There’s a high chance your waitress knows her regulars better than Google does.

Even though there’s clearly rapid development in specific cases, artificial intelligence is still just statistical prediction based on sampling and lots of data.

And as Mark Twain once nicely put it:

There are lies, damned lies and statistics.

Mirror, mirror, on the wall, which AI is the best of them all?

Companies will, therefore, hire deep learning solutions and make their decision on the most suitable one based on accuracy and cost.

Deep learning systems will indeed become one of the most useful tools in marketing, but among the various ones available, there will still be a need to choose. They will need to be trained and used with our own examples.

A game without a winner (and flavor)

Have you ever wondered who all uses Google AdWords? You and your competition?

We all contribute the same budget into the same system, which displays almost identical ads to the same target audiences?

There’s something unappealing about this future, right? Exactly, the same cappuccino for everyone.

If we leave everything to deep learning systems and employees no longer intervene, users will see the same messages at the same time, in the same place.

I can’t speak for you, but I’d prefer my cappuccino without milk.

And to conclude, here's the biggest obstacle to the dramatic conspiracy theory, where robots drink all the coffee, and humans are forced to drink water. You know, the theory that a small percentage of greedy capitalists control everything through robotic production and automated decision-making systems?

Yes. In marketing offices and departments, far too much coffee is certainly being consumed. Every month, the same Excel spreadsheets are opened, then copied into a new version, so someone can make (almost) the same decision as last month and last year.

Artificial intelligence will take over most of these tasks, or at least significantly change them. And it will open the door to new dimensions of marketing. To variations of coffee that are currently unknown, but will emerge over time as modern trends.