Kevin Hoffman's Blog
Blathering about development, dragons, and all that lies between
What AI will do to our brains
AIs impact on the human mind
An objective, research-based reflection on what AI has been and will continue to do to our minds
Our brains are remarkable things. They keep us breathing while also allowing us to ponder the secrets of the universe and enjoy memes about people getting kicked in the crotch. At their core, our brains are pattern matchers and shortcut takers. Obviously this is a simplification, but a surprising amount of how our brains work comes down to these two things.
We pattern match so that we can identify friend or foe before that stimulus hits our conscious mind. This is why we rapidly identify shapes and colors before we read the text. Our brains have already reacted to the red octagon (at least in the US) before we even start reading the word ‘Stop’ on the sign.
Shortcut-taking is our brain’s way of optimizing its own behavior to make it most suited to the activities we want to do. We routinely use the term “muscle memory” to refer to patterns that our brain has burned into shortcuts that can be taken with little or no conscious thought.
People of a certain age can remember when schools started allowing students to use calculators in class. As this became widespread, students lost their optimizations for doing math in their head. Their brains decided to make room for something else since they could now rely on a calculator to get the job done.
Evolutionarily speaking, a “short while” after that, GPS (or “SatNav” depending on where you’re from) became a ubiquitous commodity. Practically every car now has onboard navigation, and if it doesn’t, their owners' cell phones can do it. There have been a number of scientific studies that identified the area of the brain responsible for navigation and mapping. When someone is driving to a location based on a map, remembered instructions, or landmarks, that area of the brain is lit up like a fireworks display. However, when we’ve relinquished control of the mapping activity to our cars or phones, our brains are taking another shortcut. Instead of burning calories trying to manage all mapping aspects of a trip in our head, our brains can relax and passively follow the instructions of the “GPS lady” on the speaker. Often literally on autopilot, that part of our brains is a dim bulb.
Not every shortcut our brain takes is for our benefit. When driving on autopilot, so much of the brain’s normal activity becomes unnecessary that people can tune out and become dangerously oblivious to their surroundings. I’m an aberration in that I use all that spare capacity to become hypervigilant and an absurdly defensive driver.
Next up is the internet search engine and connected mobile devices. At this stage, every human has access to a vast trove of knowledge, including the words to Monty Python and the Holy Grail and the ability to play the Baby Shark song whenever we feel like it. This has truly made our lives better (narrator: But has it, though?).
This is the tipping point in how our minds are being shaped by AI. At this point, humans are actively choosing not to learn some things because the information is available whenever they need it. Our brains stop refreshing neural pathways to information we don’t need anymore because we can look it up. Have you ever noticed that something you looked up on Google a few months ago, you’ve had to re-look it up two or three times since then? It’s because your brain is optimizing and taking shortcuts, saving precious space for something important. It’s delegating the storage and retrieval of that knowledge to your phone.
AI is a tool. Human history is full of examples where the invention of a tool has had a seismic impact on the lives of everyone in the world. The discovery of fire You know there was that one caveman who was like, “Pfft. This fire thing is a fad. Losers." started a chain of events that brought us from cavemen to stock brokers.
These tools create these evolutionary lurches where their availability creates new opportunities and rapid change. Iron, bronze, steel, concrete, Play-Doh, all created huge shifts in the way humans live.
Now we’re finally at AI Here I’m referring to modern AI that includes LLMs and RAGs like Copilot, Claude, ChatGPT, and their ilk. . Like calculators and GPS, AI can do things in the blink of an AI that would cost our brains tons and tons of calories. It’s time to take more shortcuts.
It’s time to stop learning.
This isn’t hyperbole or gloom-and-doom. There now exists a huge swath of information that humans don’t need to learn. Like us forgetting something we Googled 2 days ago, there’s a bunch of things that don’t need to be retained. We don’t need to learn a new language because we can simply put an app into “translate” mode and it’ll do almost as good a job as the Babel Fish from Hitchhiker’s Guide.
Using AI as a tool implicitly declares that the destination is more important than the journey.
If you can have a computer take the journey on your behalf and simply give you the answers, your shortcut-addicted brain will salivate all over that. This isn’t inherently all bad. There are countless problems and tasks where the end result is the only thing that matters. Maybe I’ve taken the “create an empty microservice scaffold” journey 300 times and so I neither want nor need to take that journey again. I can have my AI assistant spew that stuff all day guilt-free.
When I think back on all the journeys I’ve taken where, on the surface, it might appear that I could’ve benefitted from skipping that and just going straight to the answer, I need to look below the surface. That all-day debugging and troubleshooting journey where I pounded my head on the desk until I finally figured it out: having gone through that journey has burned a level of understanding and comprehension into my mind that would never exist had I taken the shortcut.
Just like that part of my brain that shuts off while eagerly following orders from the navigation app, using AI to code will let me produce fully functioning applications that I didn’t have to think about. I will have built them in the dark, and have learned nothing from the journey. That journey went from an exercise to find my way out of the woods with nothing but duct tape and bubble gum to having a guided tour out of those woods on an ATV. How much of the woods do you think I’d remember having escaped them on my own versus the passive exit? How much more of a town’s character and layout do you remember because you rode your bike through it versus taking a bus?
Knowledge will become superficial.
Eventually, AI will pave the way for a staggering number of developers who can build applications and have absolutely no idea of how they were built or why they were designed the way they were. Once we lose our fear of hallucinations and specious results, it will get easier and easier to just blindly accept what the AI assistant tells us.
When I trust my AI assistant to summarize a report, I won’t even read the source material anymore.
Obviously there are exceptions to all of these statements. I think the key thing to remember is that AI removes the journey. For those of us who enjoy the journey (like learning languages), or those insights that can only come from having survived the journey, AI can make things measurably worse. But for those journeys I’ve already taken 300 times, I’m fine with having AI generate me a hello world microservice.
What I fear is that people like me who enjoy the journey, who see the deep value in taking the journey, and who love learning for learning’s sake, will die off like the dinosaurs. The world could be filled with people who have at best superficial knowledge on anything, experts will be gone, and AI will take all the journeys for us and only share the results.