It’s official: Google’s artificial intelligence is smarter than the competing AI from Apple, Microsoft, and Baidu…

A group of Chinese researchers put competing AI systems to the test, throughout the course of 2016.

Specifically, they were looking at the AI assistants now built into many smart phones and apps.

They ran a bunch of tests to try to test each system’s “knowledge mastery, learning, use and creation.”

And then, they correlated that with human AI scores, to try to figure out roughly the intelligence level of the different systems.

Here are the IQ scores of the different AI technologies:

— Google’s AI: 47.28

— Baidu: 32.92

— Microsoft Bing: 31.98

— Apple Siri: 23.94

I’ve gone on record saying that even complex tasks like copywriting will be able to be performed by AI “bots” within a couple decades, if not much quicker.

We’re not there yet.

In fact, these more general-use AI systems currently aren’t even as smart as a six-year-old.

But…  They ARE coming along.

In fact, from 2014 to 2016, Google’s AI nearly doubled in measured IQ.  (Moore’s Law, anyone?)

Let’s get speculative…

If they do that again, next year Google’s AI will have an IQ close to that of an average 18-year-old, or somewhere just shy of 100.

Double again over two more years (2020), and their system’s IQ will be as smart as the smartest humans, near 200.

Then suddenly, you’re into uncharted territory.  By 2024, imagine an IQ in your smart phone that’s 4X as smart as the smartest humans that have ever lived.

That doesn’t mean your phone will be able to think for itself.

But if intelligence is a measurement of your ability to acquire and apply information, this means that the smartest phones will have near instant access to all the information on the internet, and be able to gather, process, interpret, and apply that information to whatever question you ask, far better than any human being or research assistant you could ever hire.

Note, if you think this is astounding, Siri was only introduced exactly six years ago TODAY!  (Totally accidental I’m writing this article on Siri’s birthday, by the way!)

We’ve gone from having no conception of AI in our pockets to today’s tech in six years.

It’s not that big of a leap to suggest it will get substantially smarter over the next few years.

But what the heck does all of this have to do with advertising?

Well, if you make the obvious leap and say that pretty soon Siri’s going to be making ad revenue for Apple by pitching you everything from dinner recommendations to weight loss products…  You might not be too far off.

But that’s not what I wanted to talk to you about today.

Rather, I want to talk about a form of Artificial Intelligence that’s EVERYWHERE in online advertising today.

And it’s available to even the smallest, scrappiest startups.

AND you don’t even have to be a programmer or AI geek to make the most of it.

The kind of AI that was tested in the example above was a relatively generalized form of AI.

That is, when you’re creating a “personal assistant” AI service to put in consumers’ pockets, you have to be able to handle a TON of input.

Everything from needing driving directions…  To complex math problems…  To dinner recommendations…  To finding and dialing contacts in the user’s phone…  To whatever…

Not only that, you have to be able to process voice input, turn the sounds into words, the words into meaning, and THEN respond to that meaning.

Talk about complex!

Compare that to one of the most important goals in advertising:

To put a message in front of prospects who are disproportionately more likely to be ready, willing, and able to respond to an offer.

That is, to target people who are in a certain target market…  Interested in a certain topic…  Prone to recent buyer behavior…  Etc.

Once upon a time, I interviewed Brian Kurtz about this very topic, in fact.

It was his specialty at Boardroom, Inc.

He came up in the list side of the business.  First, selling lists.  Then, because he really understood lists, helping develop list selection for campaigns that would roll out to tens of millions of pieces mailed.

Here’s the thing.  It’s usually easy to mail a few hundred thousand pieces of mail, even a couple million.  (Or at least it was, in direct mail’s heyday.)

A handful of really good buyer lists, and you could hit those numbers pretty easily.

Ten million and beyond?  You usually had to get really creative with list selection.  To find audiences in unexpected places, that had a tangentially relevant connection to your target offer or creative.

For example, Brian told me how they’d look for other health book buyers, for their health books.  That’s obvious.

What’s not so obvious is that they found out that people who responded to sales pitches delivered in specific formats (e.g. magalogs) were more likely to buy from a pitch in the same format — even if the topics didn’t have an obvious overlap!  So if someone bought an investment pitch via magalog, and you had a health pitch in a magalog, that was often enough to test the list — and would often produce a winner!

That took an incredible amount of direct mail intelligence.

You had to first have a repository of direct mail experience, so you could understand the correlations between one list and another.  Then, you had to determine the important factors, and decide what seemingly-important data you could ignore in the name of testing.

Then, you had to spend a fortune ($500 per thousand pieces mailed, if not more) to test your assumptions about what might work.

Then, you’d wait weeks for response to come back, before you figured out if you were right or not.

Then, you’d probably have to run another test or two, if you thought you had a proven strategy, just to confirm your hunch before sinking big money into it.

AND, you’d have to prioritize this one test over the dozens of others you could run, on a mix of logic and gut feeling about which was the best targeting strategy to throw your limited resources at.

Today, AI from both Google and Facebook (and others) does this targeting automatically for you — on the fly, and at very little cost…

Brian sometimes jokingly calls himself T.Rex.  As in, a dinosaur.

And in terms of this one particular skill set, he’s probably more right than even he realizes.

(Although again, AI can’t touch someone like Brian for general, strategic intelligence — combined with industry wisdom — and we’ve probably got a good couple decades left of advantage over the machines!)

When you set up a Facebook campaign today, you have the option of uploading your customer list (or email list, or using a pixel to create a list of website visitors, or selecting video watchers or page post engagers, or…).

From that, you can create what Facebook calls a Lookalike Audience.

What you’re telling Facebook to do is what Brian mastered in the direct mail world.

Find other people just like my customers, who I can put my message in front of.

And here’s the thing…

Facebook (and Google, who offers a similar feature) does this automatically, and dynamically.

Automatically, meaning you don’t have to think any more about it, after you tell them to do the lookalike thing.  They run their AI algorithm, and they have a starting point for who your ideal customer is, based on far more data than you could imagine using yourself to make this decision.

But the dynamically bit is what makes it even more exciting.  Because every time someone interacts with your ad and takes a desired action, the AI takes that new data point, and adds it into the algorithm.  It’s a feedback loop.  For the first little bit, they’re operating off of assumptions.  But then, they’re testing.  On the fly.  And changing course as they go.  To put your ads in front of more and more people who they have a better and better idea are like other people who respond to that ad.

More exciting still, this algorithm is probably already operating at a level of superhuman intelligence.

You see, general artificial intelligence is constrained by having to do too much.

Very specifically-applied artificial intelligence, that solves one direct problem, can do an incredible job of moving forward on that, far faster than more generalized intelligence.

That is, the specific advertising AI is given a bunch of data about a bunch of users.

It’s told, “here are a bunch of users who displayed a certain behavior — figure out who else is likely to display that behavior.”

The AI then comes up with a set of assumptions, based on enormous amounts of data.  Then, it is constantly testing and testing and testing those assumptions.  To see which assumptions are right, which are wrong, and how to modify who it shows your ad to, to get the best possible response.

There are definitely two sides to this.  It’s frightening how this can be used, and what the AI knows.

But, it’s also an incredibly powerful marketing tool for those who learn how to tap into it.

As with all powerful marketing breakthroughs, use it for good…

Yours for bigger breakthroughs,

Roy Furr