Few topics ignite the imagination more than artificial intelligence, or AI. According to a Gallup poll, some 70% of Americans say AI will cut more jobs than it creates. Others fear much worse. To Elon Musk, it’s a technology with the capacity to enslave us, which is why he is doing all he can to stop it. He’s not alone. Bill Gates once said, “First the machines will do a lot of jobs for us and not be super intelligent. That should be positive if we manage it well. A few decades after that though the intelligence is strong enough to be a concern.” Recently the New York Times reported that AI is actively being used to create wide-scale disinformation campaigns on Facebook.
To others, AI is a technology poised to free us from the ruthless burden of toil. President Obama called it essential to improving economic productivity and raising wages. Meanwhile, in a brief titled, Economic Impacts of Artificial Intelligence, the European Parliament writes that AI will “focus on productivity, efficiency, automation and costs, enabling consumers and businesses to capitalise on the digital economy.”
Is AI coming for you in 2020?
AI already exists all around us. Every day you interact with very sophisticated AI, and even help train the AI models, though you may not be aware of your role in its proliferation. When you tell Siri to call your mom, or ask Alexa to tell you the weather, you’re engaging with a sophisticated form of AI called natural language processing. When you confirm whether or not these AI techs served up the right results, you’re helping to train the model.
It’s AI when your iPhone automatically sorts your photos or a website that suggests the perfect product you didn’t even realize you want. Websites today anticipate and serve up content, products or services most likely to intrigue us because they’re powered by algorithms designed to get to know us.
If you work for a software company, chances are high that AI is an important part of your product. And your job hasn’t been eliminated, has it? Your supervisor is a human being, not a machine, right? Now that you realize you use AI every day, perhaps you don’t find it scary after all.
What is AI?
AI is a field of computer science that teaches machines and robots to act “intelligently,” meaning they can make highly informed decisions all on their own. The concept is old, kicking around academic circles since the 1950s. In fact, a British logician and computer pioneer first described AI back in 1935.
Machine-learning is often used interchangeably with AI, but in reality it’s a subset of AI. It’s also extremely common so it’s pretty important to have an idea of what it means. Machine learning is an approach to AI where the computer learns from data (often with the help of a human).
This differs from unsupervised machine learning, which doesn’t rely on humans to identify the right answer at all. In fact, quite the opposite is true. With unsupervised machine learning, data scientists feed a machine a ton of data and rely on the machine to surface insights. This is how iPhoto makes decisions as to how to sort your photos. It looks for commonalities in images, and then files baby photos together, photos with your friends together, and so on.
Unsupervised machine learning has a wide array of applications in the field of marketing. For instance, marketers use it to predict who their next customers will be. One way to accomplish that is to look at an existing group of users and examine all of the places they’ve been and behaviors they’ve engaged in. Let’s say that you’re a store that specializes in home furnishings and you want to know when a customer is in the market for a couch long before your competitors do. There are some real advantages to that insight, namely in terms of higher conversions and lower acquisition costs. Wouldn’t it be ideal if you could advertise to those customers before they Googled ”sectional couch”? Why pay for those expensive keywords if you can target the customer proactively?
In this scenario, data scientists would look at all of the online and offline behaviors of sectional couch customers to see if patterns emerge that indicate intent. Maybe people who will ultimately buy a couch first paint their rooms — looking at interior paints online is an important clue. Machines are really good at crunching through complex and disparate datasets to identify important connections and commonalities. For instance — and this is true — people who Google school districts are likely to shop around for a new Internet service a few months later. Why? It’s likely they’re planning to move, and moving is a good time (and possibly necessary) to change carriers.
Thanks to the software industry, marketers can purchase a lot of AI functionality out-of-the-box. For instance, product recommendations and website personalization are very common tools you’ve probably interacted with as you shopped online. They anticipate what you’re likely to want or need, and present it to you automatically. Of course, these AI tools aren’t cheap, but that’s another story.
AI Depends on Humans
A big misconception about AI is that it will replace humans and eventually eliminate us, but nothing could be farther from the truth. AI simply wouldn’t work without humans. People play very important roles. First and foremost, they must frame and translate the question at hand. Let’s say you’re a marketer and you want to understand how to get more women aged 25 to 45 to buy your products. To arrive at that answer, you will need to identify the behaviors that indicate need, and then target women who share those behaviors but who aren’t yet your customers. Put another way, youwill need to translate your marketing challenge from a qualitative problem into a quantitative one that an algorithm can answer.
Once you makes that determination, the next step is to figure out if the right data is available to answer that question. Remember earlier when we talked about supervised machine learning and how the model needs a set of right and wrong answers to identify a dog? A person training AI will need to curate data in order to ensure it will correctly answer the data.
More importantly, you must train and test the algorithm, and tweak it on an ongoing basis. It’s actually not uncommon for an algorithm to go in the wrong direction if a human doesn’t watch over it. Amazon made a splash when it announced it would use AI to help it cull through resumes and decide which candidates to invite for an interview. About a year ago, however, the company had to ditch its recruitment AI because it favored only males. This is why humans need to watch over AI to ensure it’s delivering answers that make sense. A human recruiter would have spotted this bias in half a day.