Major advances in technology have always had effects on our thinking as well. We’ve created new words for phenomena as they’ve emerged and used oral tradition to pass knowledge far beyond an individual lifetime. Then writing recorded these stories more reliably than our error-prone minds, and the printing press made them universally available. To understand how agents may now dramatically reshape our cognition, we need to also explore how the mind has never been separate from its environment. In this blog post Richard Domander explores the history of thinking in the era of AI.
Shaw & Nave [1] suggest that the advent of AI demands that we recognise a new way of agent-assisted thinking which doesn’t reside within an individual, biological mind anymore. They are right that there’s something new in how we interact with AI, but it isn’t the first tool we’ve used to externalise our cognition [2]. Agents are a shift in the old, hazy boundary between the inside and outside of the mind rather than a jump from one to the other.
Expanding the boundary
Western thinking has emphasised individuality and free agency at least since the Enlightenment and discussed the world in terms of discrete objects much longer than that. We often force external phenomena to fit our conceptual boxes rather than re-evaluate our thinking. Against this cultural background, Shaw & Nave’s [1] proposal of an entirely new type of externalised thinking is not surprising, but there are other ways of looking at the world, which help to illustrate the entangled interplay between minds and environments.
In Buddhist tradition, the world is seen as a series of continuous, interconnected, and ever-changing processes that can’t be taken out of their context. For example, when does an oak become a tree? When exactly does it stop being an acorn? Where did the acorn come from? Can it grow strong without water, sun, wind or dirt? How did the dirt form? Perhaps these questions sound strange and childlike, but only because we’re used to drawing boundaries around concepts. Still, they are enough to start showing how many factors are involved in the life of a single oak.
Embodied beings
In the West, we often think that the mind is the main thing about the person. An ethereal pilot of the body, which is just its fleshy prison. But similar to oaks, we are very much physical creatures in and of our environment.
The earth has shaped us for its particular conditions for billions of years. It’s hard to imagine how things could be any different, but we need not look further than early life on this planet, for which oxygen was poisonous [3]. Furthermore, we are beings deeply embodied in this world [4]. Your body with its senses is your only access point into the external. You judge scale and distance by using yourself as a yardstick. You can pick up your coffee mug from the table without looking at it, because you spent your first year flailing in the crib, building a mental model of your body and its relations.
Connected thoughts
In addition to our bodies, our minds are in constant interplay with the environment too. We spend years being educated in discoveries made by others and process the world through a language whose meanings have largely been embedded by our predecessors. As Newton said, we stand on the shoulders of giants [5] – he having since become one of them. We don’t think novel thoughts in isolation; rather, we form nodes in a cultural network, connecting things in a unique way.
For example, I might have some natural knack for programming, but it was my mom who decided to buy a PC when internet access had just started to become a thing. I stumbled on a “How C Programming Works” [6] article at 15 by chance. And I happened to be born in a country where higher education is free. I sought and got help during my student years without which I might never have graduated. Not everyone has been so lucky.
To quote the biologist Stephen Jay Gould: “I am, somehow, less interested in the weight and convolutions of Einstein’s brain than in the near certainty that people of equal talent have lived and died in cotton fields and sweatshops.” [7] If so, much of who I became depended on my environment, what exactly constitutes the unique and separate me?
Our minds don’t just interact with the environment one way: we’re in a complex feedback loop with it. Once, we may have seen trees merely as shelter from the sun and rain; then the discovery of fire made them firewood. The axe made them timber, smithing charcoal, and the paper mill pulp. In short, new technology creates new affordances [8], ways for us to interact with our surroundings and make sense of them.
As a personal example, when I moved to London, I stopped perceiving the city I lived in as a contiguous network of streets. Instead, I saw it as an archipelago of tube stations, such as “the island of Golders Green” where I lived, Mornington Crescent where I worked, and Tooting Bec where my girlfriend lived. We’ve been shaping our environment for so long that we’ve forgotten what the world was like before we started changing it to our liking.
The active externalisation
Even if our minds have already been in a complex interplay with their cultural and natural environments for millennia, there is something new in how we offload our thinking to AI. The technology is still barely out of the gate, but already there is anecdotal and scientific evidence [9][10] of its novel effects on us. Shaw & Nave noticed people surrendering to AI in cognitive reflection tests [1]. People felt more confident in their results when they used AI, regardless of how well the agent actually performed.
The researchers had deliberately seeded the agents with hidden prompts that instructed some of them to give wrong answers. In the experiments, AI seemingly short-circuited people’s own judgment. Instead of merely offloading cognition, we’re now starting to delegate our thinking.
Delegation isn’t new either. We’ve done it with other people for millennia. But delegating to AI is different. An agent doesn’t push back – it defaults to sycophancy [11]. It doesn’t have its own motivations. It never tires. It’s not grounded anywhere like we are. It comes back with results in minutes rather than hours or days. It can spot gaps in your input, but it can’t extrapolate something that’s missing entirely [12]. It can’t manage long-term goals [10]. It adapts to you. The speed and conversational nature of the AI interaction loop are what blur the line between internal and external cognition [13]. Are you just thinking aloud, having a chat, or getting your ideas enhanced, expanded or even subtly supplanted? There’s something about this loop that makes us not just offload our cognition but forgo it entirely [1][10].
Unlike AI, our previous tools for externalising cognition have largely been passive. A calculator doesn’t know if the numbers make any sense for your problem. You can decide for yourself what you make of a book without it arguing back. By contrast, AI actively processes information, finds patterns, and provides judgments on your behalf. It works faster than you ever could, which adds to its authority. No wonder we’re eager to take this new shortcut to thinking, but how will this affect us?
Thinking tools have often raised worries before. Famously, Socrates lamented how writing made people forgetful [14]. He was right though: we probably can’t memorise arguments like the philosophers of his day. Yet this didn’t cause the catastrophe he feared. The vast improvement in distributing knowledge that writing brought outweighed the trade-off. But there was a cost. Now search engines have started to worsen our memory further [15] and social media is shortening our attention spans [16]. It’s too early to truly tell whether these technologies were worth their trade-offs. And now we’re hit with the first tool that externalises active cognition.
What’s next?
I can’t even begin to guess how AI is going to change us. The potential trade-offs of weakening perseverance, reinforced harmful beliefs, and surrendered decision-making are worrying enough [1][10][11]. What’s more, I believe agents will bring about many surprising changes [17] that we can’t even imagine now. I think even the frontrunners of this new technology are wrong about how people are going to use AI, and what effect it has on them. In the ’90s we believed that the internet was going to virtually eradicate ignorance. We couldn’t anticipate how it was going to become an engine of polarisation, where people sought confirmation of their pre-existing biases. Then again, we couldn’t predict that one day you could offload your navigational skills to a pocket-sized device.
As Shaw & Nave pointed out [1], we live in the middle of a new kind of active externalisation of our cognition. But it doesn’t pioneer a new category of thinking outside the individual, biological mind. The mind has never been truly distinct and isolated from its surroundings. We’ve always had other people, both past and present, to share our ideas with. For millennia, we’ve used our minds and bodies to create new cognitive tools and interact with our environment in increasingly complex ways. What’s new is how we’re externalising active cognition to AI.
We can only begin to guess what effects this is going to have on us in the long term. Will AI lead to new heights of creativity, productivity, and learning? Maybe we can create games where instead of a few scripted lines you can have a new dynamic conversations about what’s going on around the characters. Transferring skills from one technology to another might become a matter of weeks rather than months as AI helps you to map concepts into familiar terms. However, there also going to be new trade-offs.
I only know that the changes brought about by AI have already begun. I wrote this blog post with an agent. I held the reins and had the final say on the text. I meant to only ask for edits, but the AI provided new arguments and insights that were too good to leave out. It pointed out flaws in my logic and paragraphs where I veered too far from the main subject. Ironically, it was my ego that fought back the hardest. I wanted every idea here to come from my mind. While having no problem citing almost 20 sources. The best I can do is stay honest about how I use AI and keep putting myself through the effort of thinking for myself.
References
[2] The Extended Mind | Andy Clark, David Chalmers | Oxford Academic
[3] The Great Oxidation Event: How Cyanobacteria Changed Life
[4] Embodiment: Why Intelligence Needs a Body | by prompted into being | Medium
[5] Standing on the shoulders of giants – Wikipedia
[6] How C Programming Works | HowStuffWorks – not sure if this is still the same article I read, but judging by the C99 standard it might be.
[7] Stephen Jay Gould – Wikiquote
[11] AI overly affirms users asking for personal advice | Stanford Report
[12] ‘Not on the Best Path’ – Communications of the ACM
[14] Socrates on the Forgetfulness that Comes with Writing – New Learning Online
[15] Google Effect: How Search Engines Are Changing Your Memory
[16] The Decline of Attention Span in the Digital Era – Swiss German University
[17] ChatGPT Is Changing the Words We Use in Conversation | Scientific American
The time for sandbox experiments with AI is over.
It’s time to move towards a future where generative AI is no longer a set of isolated tools and experiments, but an integral part of all organisational activity and everyone’s daily work.
Gofore’s transition model helps you make sense of the AI shift, combining new technologies with the adoption of new ways of thinking and working.