Surely You’re Not Thinking, Mr. AI!
Written by Koen van der Pool, edited by Afaan Muhammad and Marc Bouwman
It was during a philosophy class that I first heard about the extended mind thesis—in simple terms, it’s the idea that cognitive processes can take place outside your brain and into the environment, using tools, objects, and even other people [1]. In an age where artificial intelligence (AI) lies at the heart of many tools and is becoming an everyday companion, it offers an example—and a potential challenge—to the theory. This idea instantly captured my attention.
Philosophers Andy Clark and David Chalmers, who developed this theory, illustrated it by introducing a fictional character named Otto. Otto, a man with memory loss, relies on directions written in a notebook to find his way around, making it a part of his memory. At the time, this idea challenged my worldview, as I was a firm believer in the “we are our brain” concept; the notion of cognition existing outside our body felt alien to me. Over time, however, I began to see the elegance of the theory and how it explains our tendency to delegate tasks to external objects.
Whether these processes truly happen outside our skull, I’ll leave that to the philosophers, psychologists, and neurologists to answer. However, I would argue that many tools are undeniably helpful in my thinking, allowing me to focus or prioritise. For example, as someone who likes clarity and overview, I often use an agenda to organize my tasks. Once planned, I can simply look at my schedule to determine what to do next—no need to commit everything to memory.
Feynman’s Lesson
When done well, integrating the external world into your cognitive processes can enrich your thinking. This idea reminds me of a passage from Nobel Prize-winning physicist Richard Feynman’s autobiography, Surely You’re Joking, Mr. Feynman! [2], where he, likely unintentionally, gave a clear example of how a tool can extend the mind. While attending Princeton, Feynman took a biology class and was surprised how many interesting questions remained unanswered. For an assignment, he was tasked with understanding certain functional relationships between nerves and muscles in a cat’s body. Not knowing the location of certain muscles, he consulted a zoological chart. During his presentation, he drew the chart on the board but the class interrupted him, claiming that they already knew this by heart. Feynman responded wittily: "Oh," I said, "you do? Then no wonder I can catch up with you so fast after you've had four years of biology."
Feynman's point was simple: the students were wasting time memorizing the chart—something he could look up in fifteen minutes—while he focused on understanding how the body actually worked. His approach illustrates how tools can save cognitive effort and free up mental resources for deeper, more meaningful understanding.
Feynman doing what he does best: making the complex seem simple.
You might be wondering where I am going with this. Wasn’t this going to be a post on AI? We’ll get there, as all these thoughts made me wonder: When do tools and objects extend, and when do they replace the mind?
Having It All in Your Pocket
Clark and Chalmers wrote their thesis in 1998, and Feynman walked the corridors of Princeton in the 1940s, so it is safe to say that the kinds of tools and objects around us have changed drastically. Fast forward to today: everyone’s carrying an iPhone. With smartphones, many simple tools were combined into a device that fits in your pocket: a notebook, a planner, a calculator. But more powerful tools also became instantly accessible.
It’s easy to forget, but with a smartphone in your pocket, you also had the internet at your fingertips—this was a game-changer. My first smartphone with internet access was a BlackBerry Bold 9700. I loved it; I could message my friends at all times, and if I wanted to know something, I’d just Google it in a minute. And, of course, every new technology has been accompanied by criticism—and rightfully so. In 2008, journalist Nicholas Carr voiced his concerns about search engines and the internet in his piece for The Atlantic: “Is Google Making Us Stupid?” [3]. Despite Carr calling the internet a gift for him as a writer, he also observed a trend: the internet was making it harder for us to concentrate, encouraging us to skim through information rather than engage in deep reading.
As someone born in the late nineties who grew up with computers, smartphones, and Google as the homepage of my browser, I would be lying if I said I experienced the same thing as Carr did back then. Skimming the internet to quickly determine whether a source is valid and valuable is something I learned in school. So, where Carr felt something negative shifting in his way of thinking and openly questioned the course of this growing dependency, I can’t imagine a world where we aren’t reliant on tools such as the internet. They truly feel like an extension. But more than 15 years later, as I read Carr's work now, many of his feelings about our thinking resonate deeply with me as well—particularly with the emergence of more interactive tools like ChatGPT.
A Double-edged Sword
When OpenAI launched ChatGPT in 2022, I was blown away. I saw it as the evolution of Google—you could ask anything you wanted and get an instant answer. At the time, I was working on my master’s in AI, so this was all my fellow students and I could talk about. We experimented endlessly, trying tricks to see what it could do and where it would fail. And with each new version, the bar was raised, and other major tech companies launched their own models. Suddenly, it was everywhere, all the time. It was incredibly exciting. But with all new things, you get used to it. So, over time ChatGPT became a part of how I work.
However, this is where the promise of tools like ChatGPT blurs with its pitfalls. While it extends our capabilities, it risks replacing the effort and creativity inherent to human thought. (The old) Google differs fundamentally from current language models in one important respect: interactivity. With a search engine in Carr's day, you typed in a few keywords and had to scroll through links to find answers for yourself. ChatGPT, however, engages. It is capable of helping you think, plan, and create. Already with just a few vague bullet points, you can have an email written for you.
ChatGPT launched on 30th of November 2022, with an everlasting impact.
For many instances, this has been a godsend; a tool that helps me to focus. Most of the time, I use it as an expert to help me better understand concepts or provide feedback, and I’ve gained clarity on countless ideas as a result. But human nature has its pitfalls—laziness being a big one—and ChatGPT is all too good at enabling it. This is where I connect with Carr's concerns in the present day. Although entirely anecdotal, I’ve noticed changes in my thinking. In short: I’ve become worse with language. When I want to make a point, it often sounds more like a series of disconnected thoughts, and I find it harder to get to the core when writing it down. Even for this text, ChatGPT was occasionally consulted to structure my thoughts and refine my sentences. At one point I started using it for silly tasks— checking my WhatsApp messages with ChatGPT before sending them. That’s when I realized I was on the wrong path. Every time I rely on a language model to finish or improve something, it feels like outsourcing my thinking, making me feel lazy. And this brings me back to my original point: using the tool this way doesn’t feel like an extension—it feels like replacement.
A Hybrid Future
It makes me wonder what a man like Feynman, who cherished the process of discovery, would think of ChatGPT. Like Carr, he might question whether we’re truly extending our minds—or merely outsourcing them. The answer, perhaps, lies in how we choose to use these tools and, with that in mind, how we design them. Taking this as an opportunity to reflect and act would be the desired, positive approach—because let’s be honest, there are so many applications of AI that offer tremendous benefits. It’s the reason I chose to study it and have dedicated my professional life to it. Every other day something passes on my feed that makes me incredibly excited. A great example is Khanmigo, a personal tutor from Khan Academy. If students need extra clarification on a lesson or have a question during an assignment, they can ask this assistant without being handed ready-made answers. This way, everyone can receive the help that suits them while still being challenged to think for themselves. A Richard Feynman in your pocket, who wouldn’t want that?
“The question is not wether these instruments will shape our future as they undoubtedly will. The question is how we will choose to shape theirs. ”
Tools like these show that AI has the potential to provide more than just convenience, it can act as a dedicated helpful partner. But these partnerships ask for a shift in how we think about tools and how they are related to our cognition. A challenge we simply have to deal with as the traditional thesis never had to deal with extensions that might disagree with you or take unexpected actions. Given the distinct qualities of humans and machines, there lies great potential to develop a hybrid form of intelligence; systems that complement human abilities, compensate for their flaws, and ultimately lead to better outcomes. Examples of such hybrid systems can be found in education, medicine, and even space exploration. For instance, research shows that astronomical objects and events are better detected when human and machine classification are combined [4]. This demonstrates a true synergy, preserving the unique strengths of each entity.
I propose using the framework of the extended mind as a lens to evaluate whether these systems truly extend human cognition—a necessary assessment to navigate our trajectory. If AI is to fulfill its promise to truly extend our mind, we must commit to thoughtful design and intentional use. And with AI agents—software embedded in our devices, capable of autonomously executing background tasks—already poised to make their debut, there is no time to waste. The question is not whether these instruments will shape our future as they undoubtedly will. The question is how we will choose to shape theirs.
[1] Andy, C., & David, C. (1998). The extended mind. Analysis, 58(1), 7-19.
[2] Feynman, R. P. (2010). " Surely you're joking, Mr. Feynman!": adventures of a curious character. WW Norton & Company.
[3] Carr, N. (2008). ls Google making us stupid?. Atlantic Monthly, 302(1), 56.
[4] Wright, D. E., Lintott, C. J., Smartt, S. J., Smith, K. W., Fortson, L., Trouille, L., ... & Young, D. R. (2017). A transient search using combined human and machine classifications. Monthly Notices of the Royal Astronomical Society, 472(2), 1315-1323.