Transcription
If you've been using ChatBT to learn for the last year, you are probably dumber today than you were a year ago. In fact, there's a recent paper out from MIT called Your Brain on Chat GPT. And the results are alarming, but not surprising.
Here's what they did. They separated people into three groups and got them to write essays. The first group used only chat GPT and this was called the LLM group for large language model which is things like chat GPT or Deep Seek or Gemini. The second group was allowed to use any other website but AI was strictly forbidden. This was called the search engine group. And the third group used just their brain. This was the brain only group.
They then compared the brain's electrical activity levels between each group using EEG analysis. Here's what they found. People in the LLM group had significantly lower brain activity, weaker brain connectivity, and lower levels of engagement. When they tested each group on their memory and the essays they produced, the LLM group had worse information recall, and they produced poorer quality, more generic work.
But here is the really concerning part. Even after the LLM group stopped using AI, their levels still didn't return to the same as that of the other groups, which means there is a residual negative effect of using AI. And so what this study suggests is that the more you use artificial intelligence, the worse your actual intelligence becomes. And it's not just this study. There are new studies constantly popping up over the last year that show that higher AI use is linked to lower levels of critical thinking and learning ability. And so, while everyone is worried about AI coming in to steal their jobs, the real threat is that by using AI wrong, you're actually making yourself unemployable.
Now, just to be clear, I am not anti- AAI. I'm a learning coach and a former medical doctor. And for the last 2 years, I've actually been developing our own learning science AI in partnership with Google. I've studied AI and I work intimately with AI development every day. But I also work with thousands of learners who use AI believing it's helping them when it's actually sabotaging their learning. And the truth is AI is valuable, but you have to use it correctly.
So how do you get the benefit of AI without sacrificing your own intelligence? Well, to answer this, we have to understand why AI is harming us in the first place. Let's think about the process of learning. There's a lot of great research out on how learning happens. And what we know is that when you receive information and this information can come from any kind of source. When that enters into our brain, our brain has to do a series of different types of thoughts to organize this information and make it make sense. And so this is this golden step where learning actually happens called information processing. When you do this processing really well, that's when you get a good memory. When you are able to build expertise, all the things that we want when we learn are achieved by doing this processing really well.
Now, the key is that when you learn correctly, you spend a lot of time in this part. You spend a lot of time doing the processing. And this is a part that involves a high amount of effort. This is a lot of like mental exertion happening in this step. And it makes sense because to process correctly that allows you to develop a good memory and to have deep expertise on what you learned, you can't just look at the information superficially. You have to actually actively connect this. You have to compare that information. You have to think about it in relation to a big picture and you have to eventually form what we call a schema of knowledge.
Now when we think about learning through something like books or lectures or just using Google search the information is here and the information is not exactly packaged in the easiest way for your brain to process necessarily and so to make it make sense in our brain we have to go through this very active process of organizing the information and evaluating it.
Now the difference with AI and using things like chatbt is that it seems that we are able to skip this step. we can actually bypass the processing part. So instead of having to connect it and compare it and think about a big picture, we can just go straight from having the information prompting chat GPT to help organize it and make sense of it for us so that it feels easier and more intuitive and simpler to understand. And all of this creates what's called the illusion of learning. And it stems from the fact that we believe that when we understand something that we read, we have learned it. And that is absolutely not true. Just because you can understand something that you read doesn't mean that the information is going to stick into your memory or that you can use that information in the way that you need to use it. Especially if you need to use the information for things like complex problem solving.
And so what happens is that we will use chatbt and other AI in order to extract information and instead of doing all of this effortful processing, we will just prompt it again and again to make it easier to understand so that we understand it and therefore we fall into the trap of the illusion of learning. We assume that we will have a better memory of this, that we will actually be able to retain it and that we will be able to use that information with correct expertise to solve problems and make decisions. But that is not the case because it is this process in the middle that takes effort, this processing part doing these things, creating the schema that actually creates the memory and the expertise. By bypassing that process, we are also bypassing the result completely.
To make matters worse, when you get used to bypassing this process, your brain doesn't develop the habit of processing it effectively. Which means that if a new topic has been overwhelming for you and therefore you use chatbt to make it less overwhelming because you don't know how to process it yourself, that topic will always be hard for you because you will never develop the skill and the thought patterns necessary to actually organize it for yourself. And if it wasn't bad enough, to make matters worse, we now also need to think about the fact that your AI is hallucinating. If you're not familiar, hallucinating means that the AI is generating text that is not necessarily true. And so, if you don't already have expertise on a domain, you don't even know enough to be able to check it and see if it's true or false. In fact, if you have used ChatBT to do some deep learning or complex problem solving, it is almost certain that you have already learned something that was wrong.
And the reason I can say that with such certainty is because of how LLMs like ChatgBT are built. Chat GBT is not connected to some universal source of truth. It has no concept of what the truth actually is. LLMs work through a probability network. They take your context and your query. They form a probability network and then they think about what is the most high probability word that would come next in this sequence. Recently, Apple released this white paper that has been going viral and making the rounds which showed that chat GBT and other LLMs are not actually capable of advanced reasoning. It's barely capable of reasoning at all beyond the most basic level. And this is not at all surprising because an LLM only knows what is a high probability word to come next in a sentence based on its training data. And so if you are trying to learn something or do some complex decision-making or problem solving that is very very contextual and very deep and there may not be a lot of publicly available mainstream knowledge about it then what the LLM thinks is a high probability answer can be very unreliable.
On the other hand if you do have expertise on a subject you can check the answers that the AI gives you. You can write better questions. you can give it more specific prompts and the result you get from your interaction with this AI is actually going to be a much higher quality than the interaction that a amateur would have. And this is ironic because it means that tools like chatbt don't make it so that expertise is less important. It actually makes it clearer when someone does or doesn't have true expertise because now it is so much easier to get a middle-of the road generic mainstream answer.
I remember just a couple of weeks ago I was talking to one of my learners who's a programmer and a data scientist and they've been spending weeks and dozens of hours in chatbt trying to analyze this data and figure out what the best way to build this dashboard is. So I asked him to just explain to me at a higher level. Forget about what Chachip is telling you to do. Just tell me what is your approach? What's your strategy? How do you see it all fitting together? And he really really struggled. Despite having spent so long working through this data and prompting chat GBT, he still didn't have it organized in his own head how to approach the problem. And after about an hour of discussion, we were able to get a level of clarity on this that he had never had before. and he realized he didn't even need Chachi in the first place.
But the thing is, I completely get the sentiment. He's busy. He doesn't have time to balance his work and his life. He's trying to get ahead in his career. And Chachi BT just seems like the solution that can speed up his problem solving and help him make these decisions so much faster than he could before. And if he doesn't use it, he feels like he's going to fall behind. And I completely get that. But the thing is that using it in the wrong way could be the reason that you fall behind. And there are some other things that you should do if you do want to get ahead and you want to save time and you don't want to unnecessarily waste your time doing things that you know when there is a better way to do it. And one of those things that has been proven by thousands of people over the years is my newsletter which is completely free where I've distilled through with my own brain the knowledge and expertise I've gained on learning to learn over the years and I write them with my fingers to try to give you the type of transformation that they had for me. the type of things that allowed me to finish medical school and work as a doctor while running two businesses full-time simultaneously and having enough time for family and social life. The type of unlocks that allowed me to study 10 times less than everyone else and finish a full-time masters and come first in my cohort while working 60 hours a week. And so if you are busy and you're struggling to balance your work and your life and your learning and you're not really sure how to tackle things in the best way, then you might be interested in joining my newsletter. I'll leave a link in the description for you to sign up if you'd like.
I want us to think about a different question. What if this whole problem with AI and it reducing our own intelligence is only a problem right now? What if chat GBT version 10.0 0 is so good that it does give you just the truth, factual information that is perfectly organized. What if a couple of years from now, you genuinely won't even need expertise to use it correctly? You can just give it a vague, terrible prompt and it will just figure out what you want and then just deliver it perfectly for you. After all, it's not exactly a competitive strength these days to be so strong that you can pull a card and sew the fields yourself. Maybe the ability to do deeper thinking at some point will also not be necessary anymore.
Well, if AI keeps developing like it's developing now, it is actually going to become the opposite. And the reason is that AI raises the standard for everything, not the other way around. Imagine you are an employer and you're interviewing someone and they tell you, "I don't know how to do a Google search." You probably wouldn't hire them, but you go back 30 years and this is pretty common. At a time, it was impressive enough just to have books or be able to read. Now, it's a given. It used to be impressive to know how to use the internet really well. Now, it's a given. Once upon a time, having a university degree meant you would have a stable career for the rest of your life. Now, a university degree is just the beginning for a lot of professions. How much someone is willing to pay you depends on the amount of value you think that you can provide. So, if AI is able to provide a high amount of value already, then why would someone pay you? The threshold of expertise you need to be able to bring to the table will become higher than ever. This is the impact that AI will have and is already having in the professional world. And to remain competitive, you need to understand not just how to use AI to access information which everyone can do or to understand what it provides you which anyone can do. You need to know how you can use AI to develop real expertise in your brain so that when you walk into a room, you know that any AI or tool or problem that you interact with, you can produce a higher quality result than someone else.
And so how do we do that? The simple solution is to come back to this diagram here. The next time you're trying to learn something using AI, pay close attention to when you enter into this cognitive bypassing. In the research, uh some people call it cognitive offloading. Think about when you feel like it takes a lot of effort, that things are confusing, that there's a lot of information, that it feels overwhelming, that you need to take time to think through it and piece it out. That feeling is the sign that your brain is about to start learning and building that expertise. Get into the habit of thinking about AI like an assistant as opposed to your actual brain. Use it to find you resources to be a sounding board to check for different types of perspectives. Maybe if there's a huge amount of information and you just want to get a quick understanding of it so that it helps your brain think about it, you could ask it to give you a simple overview. These are acceptable ways of using AI because it doesn't take the responsibility of processing and organizing that information outside of your brain. You're not offloading this. Instead, you're using the AI to save time on all the other little things so that you can spend your time on doing that deep critical thinking.
If I'm learning something new, I might start with asking Chachi BT to give me a big picture summary of all the key ideas within this field and I'll try to build a big picture understanding in my mind and I will challenge it and I will interrogate it. Hey, you've said this, but that doesn't seem to align with this part and this part. What about this perspective? Aren't you missing this thing? My focus is not on understanding what's being generated. It's on looking for gaps, looking for connections, looking for the missed perspectives. And once I feel like I've got a good general understanding of it based on this interaction, I feel ready to ask even deeper, more specific questions. So, I'll go through deeper, more specific resources like journal articles or textbooks. I'll buy additional books and read through those. I'll do that good old-fashioned traditional normal kind of studying because I know that that is a faster way for me to build expertise in my brain that I have ownership over and that I can retain compared to just continuing on with chatbt where I'm going to reach a point where my questions are so specific and so nuanced that I can't even trust what it's telling me anyway.
So use AI to make learning less tedious, less menial. It can save you time on the things that you'd normally be spending hours like searching through and looking for different perspectives. But the important thing is that if you want to learn, then mentally the effort has to be there. Do not try to escape this effort. Otherwise, you are really just escaping your goals. And if you feel like this process is really difficult and you're overwhelmed and you've never been good at it, then that's a sign to train the skill. You can get better at it. It is just a habit of thinking that you have to develop. But avoiding that by using AI is only going to make you fall further behind. And if you want to learn a little bit more on how you can train your brain to do this kind of effective processing a little bit more, then you might want to check out this other video that I have right here. Thanks for watching. I'll see you next time.