Keeping It Human in a Digital World (9)

As we grapple with understanding this technology, even AI pioneers struggle to find the right words. One of my favorite questions for AI builders and researchers is to ask how they describe these new systems. I’ve heard metaphors ranging from “a discovery as powerful as fire or electricity” to “a prosthesis for thinking,” while others have described it lightheartedly as “a crazy uncle” or “an untrained intern.”

Mustafa Suleyman, the CEO of Microsoft AI, argues that we should start to think of AI as a new kind of digital species. “Think about what they already do,” he explains. “They communicate in our languages, they see what we see, they consume unimaginably large amounts of information. They have memory, they have personality. They have creativity. They can even reason to some extent and formulate rudimentary plans . . . so saying AI is mainly about the math or the code is like saying we humans are mainly about carbon and water. It’s true, but it completely misses the point.”

Ethan Mollick, a Wharton professor dubbed “the go-to authority” on AI by the Wall Street Journal and one of the most passionate and insightful explorers of the technology I’ve encountered, offers a compelling perspective. He suggests, “The right analogy for AI is not humans, but an alien intelligence with a distinct set of capabilities and limitations. Just because it exceeds human ability at one task doesn’t mean it can do all related work at human level. Although AIs and humans can perform some similar tasks, the underlying ‘cognitive’ processes are fundamentally different.”

Toshi Anders Hoo is the Director of the Emerging Media Lab at the Institute for the Future. He compares AI to an array of metaphors, but one that really resonates with me is “artificial intuition.” We think of intuition as “knowing” in a way that is not entirely accurate or even based on fact but is derived from vast past experience and stored knowledge. Intuition allows us to make rapid assertions or decisions—typically without a full understanding of why. Both artificial intuition and human intuition are somewhat mysterious: we don’t fully understand how either works. Yet, it is easy to imagine how human intuition, which is (currently) still superior, can benefit from the perspective of machine intuition, which is informed by more data than a human can physically ingest over a lifetime.

Minion Mode: When AI Is an Everyday Assistant

These descriptions don’t quite fit, however, when we are turning to AI to do basic planning and tasks—a role that is more like an AI assistant. As the space evolves, we’re also seeing the emergence of “AI agents” that can autonomously perform research, make decisions, and then act on these decisions. For now, AI agent use cases tend to be more administrative than strategic. For example, an agent that operates as a trip-planning assistant could book a flight and make hotel and restaurant reservations. An agent that operates as a “digital teammate” could monitor your company’s Slack channels, identify items that need to be discussed as a team, and send out calendar invites populated with an agenda and supporting documents. If you haven’t already interacted with an AI customer service agent, you likely will soon. Companies are increasingly using AI agents to get customers faster answers to simple questions while freeing human agents for more complex issues. As AI continues to develop, we’ll need new ways to describe its different roles—just as we have specific terms for different types of traditional software.

It’s Time to Invest in Building a New Relationship

Regardless of how we describe it, using AI today requires investing in a new relationship with software. This may be startling language—after all, we’re talking about bits and bytes—but the concept of a “relationship” does give us a familiar mental model to guide our learning.

When we start a relationship with another human, we invest time in learning how to interact effectively and communicate productively. With AI, while the relationship is one-sided—it’s a tool to serve you, and you don’t have to worry about its needs—it still requires patience and experimentation to establish a consistent, productive cadence.

When we invest in a new human-to-human relationship, whether it’s a coworker, romantic partner, or friend, we try various communication approaches to find what works best. We know to get to a good cadence in a relationship, there will be trial and error, and it will take time and patience. The more curious we are about the human on the other side of a relationship, the more likely we are to not only uncover the best way to connect with them but also understand their biases and blind spots. And the more effort we put into the relationship, the more value we get from it.

These same principles apply to cultivating a relationship with AI. It takes experimentation to discover the optimal way to communicate with AI. Curiosity can drive you to new breakthroughs that help you match its capabilities better to your needs. And the more you invest in understanding and working with AI, the greater your returns.

How can you begin cultivating this relationship? By honing your AI-human communication skills—which is the focus of our next chapter.

iNote that while AI is broadly accessible to those with open internet access, some governments regulate and control internet access within their borders, impacting access to certain AI systems. The utility of the models is also impacted by the language of the user. LLMs are trained on vast amounts of internet data, so this also means communities with limited internet access may be underrepresented in these datasets. However, this “language barrier” that’s been baked into LLMs extends beyond internet access restrictions. English remains the dominant language on the internet, despite only five percent of the world’s population speaking it at home. This underrepresentation of certain languages in AI training data leads to two problems. First, it leads to bias in AI systems. Secondly, it results in reduced functionality and accuracy of AI tools for speakers of non-dominant languages. This linguistic imbalance is likely to exacerbate existing inequalities in access to and benefits from the evolving digital landscape.

Notes

1 Mustafa Suleyman, “What is an AI anyway?” [Video], TED Conferences, April 2024, https://www.ted.com/talks/mustafa_suleyman_what_is_an_ai_anyway. 2 Ethan Mollick, “Superhuman?” One Useful Thing, May 7, 2024, https://www.oneusefulthing.org/p/superhuman.

12. Working on Your AI-Human Communication: From Code to Collaboration

Brace yourself: the first basic move to raise your AIQ is to learn how to “talk” with machines in new ways, and it’s going to feel a little weird.