In this solo YouTube talk, Helena Liu argues that heavy reliance on large language models like ChatGPT can impair neural connectivity, memory, and long‑term learning. She cites an MIT EEG study that compared three groups—ChatGPT‑assisted essay writers, web‑search writers, and unaided writers—and found the ChatGPT group exhibited the lowest brain connectivity and the poorest memory, even after stopping use, a phenomenon she labels "cognitive debt." Liu draws a parallel to GPS usage, which research shows reduces spatial memory, to illustrate how convenience tools can erode mental skills. She acknowledges AI’s growing role in the job market, noting that proficiency with tools like ChatGPT has become a baseline employment requirement, much like a degree once did. To mitigate the downsides, Liu proposes active learning techniques: write explanations before asking ChatGPT to critique them, use the model as a devil’s advocate, generate practice problems, and employ structured prompting (e.g., schema‑first vibe coding) to retain deep understanding. She emphasizes that learning is inherently effortful and that AI should augment—not replace—cognitive work. The video concludes with a call to action to like, subscribe, and join a free AI training course.
Chapters
Here's my hot take. ChatGPT may make you feel smarter, but it's actually making you dumber. Recently, MIT ran a study where they divided participants into 3 different groups. One group got to use LLM like ChatGPT to write essays. Now, the second group, they got to browse the web. They got to search anything up on Google that they wanted to write the essay. And finally, the last group had to use only their brains while they were writing the essays. Now, during the study, the researchers hooked all participants up into machines where they can measure all of their brain waves, and this is called EEG. And the results that came back were astonishing. The group that used ChatGPT to write essays got the lowest level of brain connectivity and neural activity going on. So basically, they weren't using their brains as they were writing. They also had the worst memory. They don't even remember what they had written. And worst of all was the residual effects. So even after they stopped using ChatGPT, their neural connectivity was still much weaker than the other participants. And this is something called cognitive debt. So the more that we use AI tools to do all of the work for us, we actually bypass our learning process. So our learning takes place by taking in the context and the situation first. We take in all the information. We don't know how to make sense of this information yet. And this is where our brain goes to work and create all of the cause and effect relationships until it becomes second nature to us. It's like tying a shoe. When you first saw the shoelaces, you have no idea what to do with them, but through trial and error and doing it again and again— I don't just mean like watching videos or reading a book about how to do it, you actually have to put it into practice— but as you practice more and more, eventually it's just second nature for you to tie your shoelaces. And the same process applies to all learning. What AI does is that it bypasses the learning by giving us the results over and over again. And so we have the illusion of understanding a topic when we really don't understand that topic. In fact, this is also not unprecedented. So before, back in the days— and I'm not saying I'm old or anything— but when I was growing up, we actually had to use physical maps to get from A to B. We had to first look at which intersection we were at, find that intersection on the map, find where we wanted to go and then map out the streets that we have to go to get from A to B. Now we can just use a GPS to find our locations. But research have concluded that GPS actually lowers our spatial memory, and the people who use GPS more always score lower on navigation tests. So it's the same concept but different tools. Now, I know a lot of people are scared of AI replacing their jobs, but I think people should also be concerned that AI is making us dumber Now, obviously, I'm not anti-AI at all because I have devoted a good chunk of my life to creating this YouTube channel and bringing you the best AI tips that I could find. But it's also important to acknowledge the damage that this technology can have on our brains. So the solution is not to cut out all AI tools, and that's just not realistic either. It's becoming a standard For employment. Overwhelmingly, we're seeing that knowing AI is becoming the standard, the minimum bar that you need to have in order to get a lot of jobs. Here's what I mean by that. Like, if we rewind time back to like the '60s or even the '50s, if you have a university degree, that's something that makes you stand out. Today, a lot of jobs, that's the minimum requirement. Or if we rewind time back to like the '90s or even early 2000s, if you know how to browse the internet, That's a way for you to stand out. But now if you go to an employer and you say you don't know how to use Google, there's no way you are going to get hired. And so for jobs that are available today, if you say you don't know how to use ChatGPT, that's going to be a detriment. A few years ago, that's something that makes you stand out. Now that's the baseline of employment. So because you do have to know how to use these tools and they are becoming more and more integral parts of our lives, we have to use it in ways that don't just dumb us down and weaken our memory muscles, here's a couple of tips that you can employ. For example, you can turn ChatGPT into your personal coach. Whenever you're looking at a topic, maybe you write your explanation first and then ask ChatGPT to critique you. Or perhaps you can put in your reasoning and then ask ChatGPT to play the devil's advocate to point out how you could actually be wrong. You could also ask ChatGPT to create more problems about the topic that you're trying to get better at. And as you do more and more of those reps, your brain will instill that memory. So the key here is that you don't want to bypass the learning. You can use ChatGPT to help you learn more effectively, but if you just rely on it as a tool to help you think and do your work, you're not going to get the best results possible. I'll use vibe coding as an example here. If you just type into any vibe coding tool, create me an app about social media management. Now, if you don't actually know how programming works or how the database works behind the means you're not going to be able to vibe code an app better than the person who understands that. So an expert vibe coder will start with a schema first. So who is accessing this social media app and what do they have to access? They plan out the whole architecture first and they put all of that architecture into the prompt instead of just saying, hey, build me a social media app that is the difference between someone who understands that topic versus someone who does not. So even though AI is this incredible tool, we have to make sure we use it in ways that don't allow us to skip the learning process. And we have to be very cognizant of this. As humans, we spent centuries optimizing for convenience, and AI is the ultimate convenience. But learning was never designed to be convenient. It was always designed to challenge you. The pain that you go through is what makes you original, resilient, and hard to replace. ChatGPT and other AI tools can carry you over that bridge seemingly smoothly, and you may not even have noticed that you crossed it. But here's the challenge. The next time an AI answer feels effortless, ask yourself, what did I actually think about this? If the honest answer is not much, I didn't actually think about it, then you skipped the learning process. Some will use AI to skip the learning process, whereas others will use AI to have a deeper understanding of the topic. We will all use AI. The question is how. Some people will use it to skip the process and watch their brains get dumber, whereas others will use it to help them master specific topics. And those are the people who will stand out. So keep your brain on the hook. Make the machine work for you, not instead of you. Thank you so much for joining me for today's video. Make sure you like and subscribe if you got anything out of that. I've also got a free AI training course that you can join as well.
