The Long Game of Expertise and How AI Might Change Things

One of the characteristics that we associate with expertise is someone with a trove of compounding knowledge that they can recall, synthesize, and apply to new situations. Referring to Warren Buffet as an exemplar, author Shane Parrish describes it as the “filing cabinet of knowledge stored in Warren Buffett’s brain.”

In Buffett’s case, he spent a lot of time reading deeply into ideas and businesses he was interested in and visited a wide range of companies for hours at a time to understand everything from production to management. To this day, he reads hundreds of annual reports for companies in a wide range of industries that he doesn’t own. All to say, that he is a voracious consumer of information.

Parrish argues that Buffet does something different from what a lot of us do: most of us consume “expiring information” whereas Buffet consumes information with a “long half-life.” Why does this matter? Because he focused on information that changed slowly over time, it allowed him to compound his knowledge.

Expiring information vs. compounding knowledge

As Parrish describes it, expiring information has a very short lifecycle. It’s marketed to you. It lacks nuance. It’s easily digestible.

It’s also delivered in vast quantities and in a myriad of ways that support our habit of looking for fast and packaged information. This is what James Gleick recognized in 2011 as “a flood, where information is cheap and plentiful and ubiquitous.” If anything, this is even more true today.

It’s not because people set out to be trivial about their efforts to learn. There are time pressures that restrict how long or how often we can investigate our interests. The sheer volume of information released every day is overwhelming, as is an equally overwhelming array of media. It’s easier. Most of all, we have become used to relying heavily on computers to hold our knowledge for us.

Compounding knowledge is associated with expertise that someone builds through consistent learning and development over time. An expert might compound their knowledge in a skillset (e.g., an analyst who has superb proficiency with spreadsheets and formulas), a profession (e.g., an engineer versed in multiple languages and technology stacks), or a competency (e.g., a consultant who coaches executives on decision making and innovation).

This kind of knowledge comes through persistence, lifelong learning, and consistent efforts over a long period of time. Rather than relying on others to pack up fast information for us, we might dig into the details ourselves and do our own thinking. In any new situation, this kind of adaptive expert has the advantage of cognitive flexibility and pattern recognition to guide them to informed decisions.

Along comes AI

While GPT stands for Generative Pre-Trained Transformer, it might be useful to retain an earlier and equally valid meaning: General Purpose Technology, as Ethan Mollick advocates. We can frame GPT as analogous to the Industrial Revolution or the Internet. He cautions that many leaders believe the main purpose of technology is efficiency, in which AI becomes a way to cut costs.

Instead, any real advantage in AI “will come from the expertise of their employees, which is needed to unlock the expertise latent in AI.” We need experts, with that deep, compounding knowledge to both build and extend capabilities that become available via AI.

AI is still new enough that we are only speculating how AI technologies will affect learning and expertise. As knowledge workers, we may face challenges such as:

• Obsolete skillsets. Some skills will be primarily AI-led, and AI will be more proficient at them. This is pretty much guaranteed to happen.
• Dependence on AI. If we were outsourcing knowledge capacity to computers before, it’s only going to get worse. Outsourcing decision making to AI is a hot topic of discussion as one of the scariest potential future scenarios.
• Biases, skewed data, false information, and malicious intents. Any information we rely on to compound our knowledge may become increasingly unreliable. As we are asked to incorporate AI into practice, there will also be ethical considerations related to data privacy/usage that we’ll need to navigate.
• Competition for human expertise. To the extent that AI will be given the scope to solve complex problems, adjudicate options, and make strategic decisions.

A way forward

While the tools are changing and more diligence will be required, experts will continue to have an important role in work and society. There are some aspects of how their relationship to knowledge building will be enhanced with AI:

1. Task automation. In any field, there are tasks that AI will be better at. There are two considerations here. This is a good thing in that there will be more time to focus on higher level tasks and creative tasks as well as innovation (and I’m sure AI can provide some inspiration), but we still need humans in the loop who understand the underlying principles and processes of those tasks.
2. Data insights. Even with domain experts, there’s so much data available that it’s a case of “we don’t know what we don’t know” sometimes. AI can process vast amounts of data quickly, providing insights that would take humans much longer to uncover.
3. More ways to experiment. Experts with deep domain knowledge are really good at seeing patterns and applying insights from one problem set to another. What AI can likely help with is powering up their ability to try “what if” scenarios, to check/verify patterns, or to extrapolate a “gut feeling" to see where it goes.
4. Speed up problem solving and analysis. One of the benefits of AI toolsets is that they can learn from exemplars. Using expert “templates” or outputs will help the expert to apply what they know more quickly, but also allow others to benefit as well. Experts from across the business should be involved in learning about models and training them.
5. Personalized learning. AI platforms can tailor content and can enhance the resources available for lifelong learning. But like Buffet, you should always seek out the source materials that you discover this way. Deep learning won’t happen if you use Generative AI as a cheat way to be an “expert” – it’s really not optimized for that.
6. Personal idea checker. Experts seek oppositional points of view as a way to check their own views and to strengthen their knowledge. Sometimes there’s no one around who can play that role. AI is primed for this – you can assign it a role or a point of view and it will advocate for it, and you can use the interactions to better test your viewpoint.
7. Ethics and shaping AI. Implementing AI responsibly and ethically is essential. We know that there are already biases and misleading traits evolving in LLMs. Experts need to be there to shape AI development.

Continue to seek new information, delve deep into subjects that interest you, and build a robust knowledge base that you can draw upon throughout your career. How you do it might be changing, but lifelong learning will continue to be a worthy commitment.

As AI continues to evolve, it will be inevitable that experts need to embrace these changes and leverage AI tools to enhance their learning and professional growth:

• Stay curious. Keep an open mind and be willing to explore new AI technologies and non-AI technologies in your learning journey. Compounding knowledge will be a life’s work of consistent and deep learning that happens in many ways – reading, interacting, observing, doing.
• Invest in learning. Take advantage of AI-powered educational platforms to stay ahead in your field and to expose you to new original resources and data that can shape your knowledge and expand the contexts and applications you consider.
• Stay adaptable. Be flexible and ready to adapt to new ways of working as AI continues to transform the ways we learn and work.

The next generation of experts will see AI as a co-intelligence in the long game of compounding knowledge.