ChatGPT won't make you dumb unless you want to
Whether ChatGPT makes you dumber or smarter depends on how you use it
ChatGPT is making us dumber is a worry that a lot of people have. And now there’s a new paper out of MIT that appears to have “proof”.
The Bad News
The paper is "Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task". And its findings include:
“LLM users consistently underperformed at neural, linguistic, and behavioural levels”
Students were asked to write essays with the help of LLMs, or Search Engines, or without any help. And their brain functions were monitored using EEGs. They found that “Brain connectivity systematically scaled down with the amount of external support: the Brain‑only group exhibited the strongest, widest‑ranging networks, Search Engine group showed intermediate engagement, and LLM assistance elicited the weakest overall coupling.”
After writing, only 17 % of ChatGPT users could quote their own sentences, versus 89 % in the brain-only group.
etc. See this article for more details.
Sounds bad.
Is It Really Bad News, Though?
If you use LLMs to write your essays, you write worse essays, you don’t remember your own essay, and you learn less.
Is that surprising?
If you use LLMs to avoid learning, you don’t learn. This shouldn’t be news, right?
Yes, it is true that students are using LLMs to cheat and get good marks without actually learning as much. And that is making them dumb in the short term.
But that is a temporary problem caused by the fact that our culture has not caught up to the technological reality. Students cheating with the help of LLMs are doing the equivalent of joining a gym and then using a robot to lift weights for them. And this might work for a few years because both colleges that give them degrees and companies that hire them haven’t yet figured out how to accurately evaluate their actually useful skills.
In the old days, the ability to write a good essay or a small standalone program was a strong predictor of intelligence and suitability for a job. But now, those can be done by idiots using ChatGPT, but college exams and company interviews haven’t caught up to this fact.
But they will. And society will adapt.
What’s the Proof That Society Will Adapt?
In 2008, The Atlantic published a very popular essay, Is Google Making Us Stupid? Today, we can assert that this did not happen—we invented CRISPR, ChatGPT, and the COVID vaccine in record time.
In the 1960s, digital calculators were invented, and there was much hand-wringing that it was making us stupid. That didn’t happen either.
We can keep going back in time. 2400 years ago, Plato was complaining that writing was making us dumb.
Conclusion?
If you’re a student, stop being an idiot: use LLMs as an aid to learning, not as a replacement for learning.
And whether you’re a student or not, it is important to keep one key thought in mind: when you use Gen AI to automate one of your tasks, it is important to focus on what else you can do with the time you saved. And what cool things Gen AI enables you to do in that time.
Specifically, Gen AI should make you more ambitious. Things that were out of reach—either because they would take too much time, or because you lacked the skills—are now suddenly within your grasp. Use this superpower to do more—for yourself, for your customers, for the world. This should be the time to embrace Gen AI to the fullest extent, to do more, achieve more.