In a recent LinkedIn survey, I asked product and software developers if and how they were making their software smarter. A surprising 57% cited A/B testing, while another 50% reported they were still swinging from decision trees.
Why are developers still solving everyday pain points with these manual, archaic processes, as opposed to employing “Little AI”? There are millions of everyday use cases for AI, where technology is empowered to learn and decide on a course of action that offers the best outcome for consumers and companies alike. The problem is that the Big AI we’re used to has a lot of challenges that make it inaccessible for developers to employ for tasks that’d benefit from everyday AI.
There are millions of everyday use cases for AI, where technology is empowered to learn and decide on a course of action that offers the best outcome for consumers and companies alike.
What we’re missing
Take this article you’re reading right now. If TechCrunch let loose a Little AI — essentially empowered machine learning — it could learn you prefer to read short, newsworthy articles in the morning and longer thought pieces at night. That learning informs a personalized home page, presenting you with bullet points upon awakening and feature stories at night — all without you having to laboriously enter your preferences or respond to pop-up surveys.
Little AI also learns that what your VC friend wants to see on their screen first thing is recent series funding announcements. A truly personalized experience is not only our expectation, it is the core component of the relationship between us and our content providers. And yet, it’s missing.
Let’s raise the stakes. There are multiple players in the split-pay space. A sprinkle of Little AI can teach a fintech provider that one consumer likes to finish paying off an item in less than six months and never wants any outstanding payment to exceed $250. It can also learn that they are open to revolving credit/product offers for an experience-related purchase above $1,500. This type of truly personalized financing enables both the consumer and merchant to benefit from a completed sale while lowering the risk of default to the credit provider.
Travel will be coming back in a big way, with more deals than ever. Little AI can jump in and learn how to make that experience far easier for consumers and far more successful for travel providers. Rather than showing consumers every single deal for every single location, it can take personal preferences into account.
If there’s a long weekend coming up, a Little AI can show you get-out-of-town deals to places with direct flights under three hours or a drive under four hours. It can prioritize nature sans Wi-Fi, or, on the other hand, facilities with top-rated spas and excellent remote work capabilities. The offers you see would be optimized to your specific, personal location factors and desires, rather than a list of best-guess scenarios.
All of these examples showcase the small ways Little AI could improve our lives and businesses in ways that we’ve mostly shied away from until now.
Moving to smarter AI
Figuring out an ideal, personalized call to action for individual users of a website or application page; the best offer to show to increase conversions; or even the best time of day and channel to connect with a community are quotidian tactical problems. Problems that can be solved faster and better with technology empowered to learn and decide on a solution in the moment, rather than relying on grinding through large, uninspiring data pools or faulty models, which thus far have been attributed to nearly 85% project failure rates.
Decision trees utilized in traditional AI are decades old and rely on humans making broad assumptions based on deterministic if/then scenarios. But people don’t always assume correctly the first time. And deterministic technology doesn’t learn as new data points are provided. Or take A/B testing — a “mature technology” working on a “guess and test” scenario — which guesses the best option to present to a consumer. By the time an A/B model is tested and put into play, weeks or even months have passed. Users are placed in overly broad buckets that reductively simplify their needs to the testing scenarios presented. It’s too late to make use of anything learned in real time; a two-week sprint won’t catch a one-day trend.
Catherine Wood recently predicted that the potential economic growth from learning solutions will be bigger than the entire value of the internet within 15-20 years to more than $30 trillion in market capitalization creation.
To unlock that value, we must accelerate AI adoption beyond power users. Big AI is self-driving cars, contact tracing and protein folding. Little AI is everything else.
So let’s recognize, normalize and empower everyday AI applications and solutions. We don’t have to be a certified auto mechanic to drive a car. Likewise, we don’t need to be or even hire a data scientist to solve our everyday pain points when the software to do so becomes mainstream, which is not nearly as hard — or as far off — as it might seem.
Little AI is practical AI
Little AI is not Asimovian AI, or AGI (artificial general intelligence), popularly discussed in late-night Clubhouse rooms. It’s practical intelligence about our businesses, goals and customers, leveraged quickly into informed, actionable decisions.
It’s time to reframe how we talk and think about AI by debunking its black-box mythology and make its individual elements more transparent and easier to use. Nix the notion that AI isn’t accessible, or isn’t for us, and take the time to understand its more simple workings. While we might not care to learn the specifics of how Tesla’s “full self-driving” AI makes it possible for your car to pick you up in the rain, there is value in learning how smaller-scale AI can be used to shorten and advance product life cycles.
There are single, simple elements of AI that are less time-consuming, more affordable and easy enough for anyone to make everyday technology tasks smarter. Little AI, instead of Big AI. It makes sense to apply Little AI that learns to small-scale projects. Little AI doesn’t require complex guidelines and the intricate bias-prevention considerations that Big AI solutions demand.
The future is software that learns and helps develop itself. We should be deploying simple AI tools to solve and improve upon everyday tactics so we can spend more time on strategic growth goals and long-term execution. By letting Little AI serve as our rapid response tacticians, we can free up a tremendous amount of our personal mindshare to focus on new ideas and strategies.
And that’s no small feat. In fact, it might give us time to come up with the next really big idea.