Generative AI - an evolutionary tale

The introduction of ChatGPT was, for many people, their first glimpse into a new level of text generation, but the techniques used in ChatGPT, Google’s Bard, Apple’s Ferret, and other such models have been around for some time. Related techniques can generate text, pictures, videos, and sound from a user’s written prompt. For text generation alone, dozens of open-source models with innumerable variations are available. They are astounding in their ability to produce the words we want them to create, but they can produce inaccurate text using confident wording (“hallucinations”).

Also, the models know the patterns and sequences of words humans use, so they produce text in those patterns and sequences but do not understand them. It’s a bit like the age-old question of whether an infinite number of monkeys on typewriters would eventually produce Shakespeare's works. Even if they did, they would not understand what they wrote.

‍

The data that AI is collecting

The data collected today for Generative AI to pump out content is conversion-centric and product-centric, not customer-centric. Why does this matter? Generative AI scrapes incomplete and fractional data to derive holistic insights and meaning. It is being assembled without a massive piece of the customer's perspective. In the iceberg idiom, today's data is above the water line.

Sure, brands know “who purchased what” and “when, where, and how it happened,” but they have no idea why because they know very little about the customer. The functional answers delivered through Generative AI lack the emotional attributes driving the most powerful aspects of customer decision-making. Think about it. What brands have access to customer interests, lifestyles, needs, desires, preferences, propensities, attributes, sentiments, satisfaction, or values? Who knows what is truly driving customer behavior or decision-making? Which brands know the names of the children, their favorite vacation destinations, or the song that makes them smile? How is Generative AI scraping that kind of detail if nobody collects it and posts it in the public sphere? Spoiler alert: Gen AI is not.

‍

The massive gap

We need a reality check. Hyperpersonal RFM remains a black hole. The "next-best" action is a guess, sometimes educated, that makes most brands happy if it converts at 4%. Behavioral intelligence leads to more educated guesses but is only fractionally more valuable than those without behavioral intelligence. Umego delivers unique characteristics and customer-centric insights that will make Gen AI more valuable across stakeholder groups. Our models are trained towards customer-centric insights and traditional omnichannel social, conversion, and product-centric insights to provide a holistic view of the customer. Our analytics engine becomes an integration hub that delivers holistic value to every customer-brand relationship.

Before you get too wrapped up in the promise of Generative AI, remember that most only see the tip of the iceberg. As a CEO of AI at Salesforce recently stated in a Generative AI whitepaper, “Data is fuel for AI - without high-quality, trusted data, it becomes ‘garbage in, garbage out.’ AI pulling from data sources that are irrelevant, unrepresentative, or incomplete can create bias, hallucinations, and toxic outputs.” We agree. That’s why we get the data directly from the customer’s mouth.

‍

‍