Making AI more human, one sticker at a time

Copy of Copy of Making AI more human, one sticker at a time.png

Unless you’re a data or machine learning scientist, chances are you don’t find yourself debating Bayesian versus frequentist methods over a warm cup of coffee (if you do, hats off to you). The point is: the mechanics of AI can be complicated.

That’s why education is so important to us. Whether we’re talking to new hires or prospective customers, it’s critical that we effectively communicate who we are as a company, what we stand for, and what we do.

And keeping things simple is a big part of that.

In the spirit of simplicity, our machine learning scientists put together a list of some of the most enigmatic-sounding algorithms, problems, and theorems. Then, we created graphics to represent each of them. By putting a face to concepts that are typically considered abstract, we hope to make them easier to understand. We also hope to humanize them. 

For us, this is important. There’s this misconception that AI poses a threat. Mostly, it comes from a fundamental misunderstanding of what AI does, and how it should be applied (that is, responsibly). We believe artificial intelligence represents an unprecedented opportunity for us to become more human. 

Now, our boardrooms take their names from machine learning concepts, and in each hangs a framed portrait of that concept. Want to see what they all look like and find out what they mean? Keep scrolling.

 
Random Forest   Chances are you've heard of Random Forest algorithms. They’re everywhere—in diagnostic tools that identify diseases based on someone’s medical records and recommendation systems you might find on Amazon or Netflix.  The algorithm takes its name from the many decision trees that make it up. Each tree classifies a new object from an input. The classification it generates counts as a vote for that class. Whichever classification registers the most votes gets chosen by the forest.

Random Forest

Chances are you've heard of Random Forest algorithms. They’re everywhere—in diagnostic tools that identify diseases based on someone’s medical records and recommendation systems you might find on Amazon or Netflix.

The algorithm takes its name from the many decision trees that make it up. Each tree classifies a new object from an input. The classification it generates counts as a vote for that class. Whichever classification registers the most votes gets chosen by the forest.

 
Black Swan Paradox   If you’re thinking the Black Swan Paradox has something to do with the Natalie Portman/Mila Kunis movie, you’re wrong.  Black Swans are random, seemingly unexpected phenomenon that have an almost disproportionate impact. In the world of artificial intelligence, the context that’s most relevant to us, they’re data points that lie far beyond the anticipated range for a value. We could call them outliers. That’s why they’re something we need to be concerned about and address.

Black Swan Paradox

If you’re thinking the Black Swan Paradox has something to do with the Natalie Portman/Mila Kunis movie, you’re wrong.

Black Swans are random, seemingly unexpected phenomenon that have an almost disproportionate impact. In the world of artificial intelligence, the context that’s most relevant to us, they’re data points that lie far beyond the anticipated range for a value. We could call them outliers. That’s why they’re something we need to be concerned about and address.

 
Multi-armed Bandit Algorithm   This algorithm takes its name from slot machines, or one-armed bandits, as we call them. Now, think of a slot machine with a number of different arms. The problem you might find yourself facing is: Which arm to pull? In other words, which option will provide the highest returns given your initial investment? That’s the problem the Multi-armed Bandit Algorithm solves. It identifies where limited resources should be allocated to maximize returns.

Multi-armed Bandit Algorithm

This algorithm takes its name from slot machines, or one-armed bandits, as we call them. Now, think of a slot machine with a number of different arms. The problem you might find yourself facing is: Which arm to pull? In other words, which option will provide the highest returns given your initial investment? That’s the problem the Multi-armed Bandit Algorithm solves. It identifies where limited resources should be allocated to maximize returns.

 
The Knapsack Problem   Imagine a knapsack. Now, imagine you have a ton of textbooks to put in said knapsack and they all have different levels of importance. Obviously, you want to get as many of those important textbooks packed away as possible. This illustrates exactly what the knapsack problem is all about, i.e. fitting as many items as possible inside an object while maximizing the value of all items. We call this combinatorial optimization.

The Knapsack Problem

Imagine a knapsack. Now, imagine you have a ton of textbooks to put in said knapsack and they all have different levels of importance. Obviously, you want to get as many of those important textbooks packed away as possible. This illustrates exactly what the knapsack problem is all about, i.e. fitting as many items as possible inside an object while maximizing the value of all items. We call this combinatorial optimization.


No Free Lunch Theorem   Basically, the No Free Lunch Theorem states that no solution offers a short cut over any others for a particular class of problems.

No Free Lunch Theorem

Basically, the No Free Lunch Theorem states that no solution offers a short cut over any others for a particular class of problems.

If you happen to spot us at a conference or networking event, be sure to stop by. We’ve printed a ton of these graphics in sticker form, and we’d love to share them with you. Then, maybe you can share them with the world, and help us humanize AI!

 

integrate.ai is an AI-powered customer intelligence platform that helps consumer businesses make precise predictions about customers needs so they can deliver more meaningful and relevant digital interactions that drive better business outcomes. Our SaaS platform integrates with existing systems and then uses privacy-preserving AI techniques to enrich customer data sources with insights from other industries. This paints a richer, more complete picture of the customer—without sacrificing consumer privacy or security.

Previous

AI privacy and customer experience: avoiding the creep factor

Next

The AI dilemma: build versus buy