Building with a Purpose: How Businesses Use AI for Scalability and Success

What if I told you that we’re already living amongst artificial superhuman intelligence? Well, it’s true! Although we don’t have robots making our breakfast each morning (yet), we are living and working with AI. In this post, I will prove this seemingly outrageous claim because piecing it together will be helpful in understanding the current state of AI and how best to leverage it for your business.

History: From Handheld Tools to Artificial Intelligence

In the beginning, humans built tools to better harness their physical power. While spears and shovels allowed humans to apply their strength in more specialized and effective ways, even these tools were limited by our body’s physical strength and those of the livestock. As a result, humans replaced hand tools with engines and tractors, no longer limiting physical power by biology.

Intellectual progress has followed a similar trajectory. The first tool was language, which was followed by writing. Eventually, specific notation techniques were invented to make intellectual processing more efficient, like the tedious (but effective) process of long division or carrying-the-one trick for multiplication. These notation techniques are the reasons we moved from Roman Numerals like “XIV” to Arabic Numerals like “14” (just try to multiply two Roman Numerals together and you’ll see what I mean).  

Once more complex and effective tools like abacuses were created, we found that the speed of calculation was still limited by our human brains, leading to the invention of calculators and computers. In fact, Apple’s iPhone 7 can add two numbers together over a billion times a second (mind-blowing, I know), no longer limiting our intellectual power by biology.

I know what you’re thinking – tractors have superhuman strength and calculators have superhuman arithmetic power, but neither of these satisfies the claim of “superhuman intelligence.” Intelligence is defined as “the ability to learn or understand or deal with new situations,” and it’s clear that a tractor or a calculator cannot actually learn. No matter how many numbers you multiply with a calculator or holes you dig with a tractor, you don’t end up with a better calculator or a better tractor. Learning from prior experiences is a key component to intelligence and the exact goal of machine learning systems, which is what makes machine learning differ from traditional computer programs.

We’ve had computer applications that learn from experience since 1959 when Arthur Samuel wrote a program that learned to play Checkers, so now the question is: How do we measure intelligence and compare it to human-level intelligence? Usually, we give a machine and a human the same goal, like playing Chess for example, then allow the human and the AI to compete. There have been many specific goals that AI’s have beat humans at, including: chess (built at IBM in 1997), Jeopardy (built at IBM in 2011), data analysis (built at MIT in 2015), Go (built at Google in 2016), and poker (built at Carnegie Mellon in 2017).

Living and Working with Artificial Intelligence Today

When learning machines, aka AI, beat humans at specific goals, it proves that we are living with superhuman intelligence in our midst. However, human intelligence differs from AI in that we are able to handle an incredible range of goals, from general to specific, from learning languages to flying planes to playing baseball. In contrast, all of our AI systems today are only capable of learning specific goals, like how to beat a human in chess. Even the best chess AI in the world doesn’t know how to play checkers.

Understanding the breadth of goals limitations in AI is critically important for clients, vendors, and investors alike. Since AIs are only successful at tackling a specific goal, it’s important to understand exactly what task the AI was built and trained for as well as how well that training relates to the application the AI is used for.

Today, the industry is rife with misleading claims about the impact AI has on their businesses, all of which mirror some form of: “This AI does ‘A’ which makes us so much better at ‘B’.”  This claim is common because many companies don’t invest in the engineering required to build and train an AI for a specific use case. Rather, they are using an off-the-shelf AI service and applying it to their similar, but different, use case with mediocre effectiveness.

The key to understanding the value of any given AI is to understand its goal alignment by asking a few questions. What does the AI predict? What data was the AI trained on? How are these predictions used? AI that provides scalability and true performance benefits to the business are built in a particular way – with narrow goals, trained on relevant data, and ultimately predicting the outcomes of that narrow goal.  

Linqia’s AI is Built for Influencer Marketing

Linqia Performance Platform has evolved over the past six years and today we can confidently claim, “This AI was purpose-built to predict influencer outcomes for specific campaigns.” See the difference between our claim and the misleading claim mentioned above?

Historically, Linqia’s PerformanceMatch was trained on broad categories of topics, using image recognition, natural language processing, and sentiment analysis to categorize influencers and their content. But we wanted more from our AI, so we invested in engineering and built PerformanceMatch to predict and score every influencer based on how well they would perform for a specific campaign, making our recommendations even more powerful. This predictive modeling is what we use to fill every Linqia program while guaranteeing the results.

Linqia PerformanceMatch uses machine learning to surface influencers who meet the specific requirements outlined in each program’s creative brief. These requirements include influencer expertise/style, historical affinity for a brand, audience engagement rates, and click-through-rates, audience sentiment around a specific topic, topic alignment with the brand’s objectives, and competitive restrictions.

Linqia’s AI can do all of these things because we’ve trained it to learn from the six years of performance-based campaigns. For more about AI goal alignment and its impact on business, check out my previous post (originally published in Adweek) on how to tell whether AI is marketing buzz or real value.  

April did not disappoint in the world of influencer marketing and the creator economy.