The Future of Work: How Centaurs are Redefining Productivity and Success

The emergence of advanced artificial intelligence (AI) technologies, such as Chat GPT, has led to many recent discussions in my professional and social circles about the future of work. With AI capabilities advancing at an unprecedented pace, many of my friends, peers, and clients are wondering what the workforce will look like in 5-10 years. Which industries will see massive contraction and which jobs will be displaced? Will artists' jobs be replaced by AI-generated content? Will "content generation" roles, such as copywriting, software development, or report collation, disappear? Especially considering recent headlines of ChatGPT getting top marks in technical screens and passing exams for medical licensure, a Wharton MBA exam, and the legal Bar, it's easy to understand the root of the anxiety. But while these are valid concerns, the reality is that the relationship between humans and AI is much more complex than simply one replacing the other.

Interestingly, I recently finished "After On," by Rob Reid. I won't spoil the book for you as it was a fun read, but it's basically a "Black Mirror" adjacent fictional take on Silicon Valley, emerging tech, and - above all - artificial intelligence. As you might imagine, a book like this bleeds in and out of science fiction and reality, and one concept I had never heard of before but that I found fascinating is the idea of "centaurs." Centaurs are the combination of AI and human intelligence to create a system that is stronger than either the strongest humans alone or the best AI programs alone. Initially studied in chess in 1998 by Kasparov, a grandmaster who had been beaten by a hyper-tuned chess AI tool the year before, this idea has since been proven in various fields. By leveraging the strengths of both humans and AI, centaurs have been shown to be more effective than either alone.

In 1997, world champion Garry Kasparov was defeated by the computer program Deep Blue. However, in more recent years, centaur chess teams (also referred to freestyle or advanced chess) consisting of a human player and a computer program have been able to crush some of the strongest AI programs and some of the top grandmasters alone. This is because the human player is able to bring their intuition and pattern recognition skills to the table, while the computer program can analyze a larger number of potential moves. Chess is inherently a psychological game, and computers struggle to anticipate suboptimal "gambles" that humans might make. Alternatively, people acting under pressure may make blunders that a computer is simply incapable of making (or anticipating) meaning they may not always be able to accurately predict their human opponent's move with the intuition a human might.

Another example of centaurs in practice is in the field of medicine. In a 2022 study, a centaur system (consisting of a human radiologist and an AI program) was able to diagnose lung cancer more accurately than either the human or the AI alone. The human radiologist was able to bring their medical expertise and experience to the table, while the AI was able to analyze a larger number of images and detect patterns that the human may have missed.

A final example is in the field of finance. Centaur systems are able to make more profitable trades than either the human or the AI alone. The human trader can bring their market knowledge and emotional intelligence to the table, while the AI is able to analyze much more market data and identify patterns that the human may have missed.

These examples illustrate the power of centaur systems to outperform humans and AI alone. This is because centaur systems are able to leverage the strengths of both humans and AI to make better decisions and solve problems more effectively. Additionally, centaur systems can also help to overcome the limitations of both humans and AI, such as the human's ability to process large amounts of data and the AI's lack of intuition and emotional intelligence.

So just equip your best contributors with AI and watch them run? Not so fast. Kasparov, the world champion chess player mentioned above, shared in a 2017 podcast that when it comes to utilizing AI, having a strong process is more important than being a domain expert. He states that "a weak human player plus machine plus a better process is superior, not only to a very powerful machine, but most remarkably, to a strong human player plus machine plus an inferior process." This highlights the importance of not just having specialized individuals, but also having a strong process in place to effectively utilize AI, and that the process may even be more important than the specialization.

One of the key takeaways then is that the best action to future-proofing your role (or business) with AI is potentially different than in the past. In the past, the common approach was to hone your craft, domain knowledge, and expertise, then scale your efficiency and productivity (process) to maximize impact. In the new world we are heading into, the approach is to hone your process and usage of AI, then over time perfect your craft and domain knowledge.

This is an important point to consider for professionals and businesses that want to stay competitive in the age of AI. For example, if you are a developer, the skills you need to learn and focus on might be different than they were in the past. Instead of focusing on becoming a domain expert in a particular programming language or technology, you may need to focus more on understanding how to use AI tools and how to integrate them into your development process. This is a subtle but important shift in what we may want to look for in a developer, and the same may be true for other "content generators" - artists, copywriters, authors, political speech ghostwriters, marketers, report builders, musicians, publicists, etc.

In line with this shift, I had an interesting conversation with a peer, Jon Molendorp, about how to evaluate the "developer of the future." His concern about only focusing on "AI masters" without good understanding on the "why" behind the development was the proliferation of "cargo cult" programmers. These are people who can spit out scripts and code (commonly using Stack Overflow to an insane level) without understanding the why behind the what, making their resilience to change a weakness. He basically argued (correct me if I mischaracterize you Jon) that craft and process are inextricably linked, and trying to create a process without solid craft fundamentals is likely to set the engineer up for future failure. My opinion is that failure isn't inevitable, but that tech leads, architects, and engineering managers need to build guardrails (be that craft education of the operators or building resiliency into the "centaur process frameworks") to support these AI-powered individuals.

The world is changing, and those that were "the best" are not necessarily going to be "the best" in the future. What we look for in resources, how we screen them, how we support and tool them, and how we incentivize them all stand to be shaken up and reconsidered in light of this centaur idea. Let me know your thoughts on how to better evaluate centaurs in your fields, I'd love to hear other perspectives!