Episode # 97 | Is ChatGPT Making Us Dumb?

Summary

In this conversation, Jenny Chen explores the complex implications of ChatGPT on intelligence and education. She emphasizes the need for careful consideration of the technology's impact, particularly in academic settings, and discusses the broader context of a technological revolution that may have both positive and negative consequences.


Takeaways

• Is ChatGPT making us dumb?
• This is a much more complicated question.
• ChatGPT is problematic, especially for students.
• We should consider this carefully and thoughtfully.
• We're at a juncture of technological revolution.
• Revolutions typically will have some casualties.
• How can we use this new tool effectively?
• Maximize our productivity without diminishing humanity.
• Not causing something unthinkable, like being ruled by AI.
• The good and evil of this very potent tool.

Show Notes: 

https://3dheals.com/is-chatgtp-or-ai-making-us-stupid-two-cents/

Transcript

Monologue Debut & Theme

SPEAKER_00 0:00

Hello, hello, hello! Welcome to the Ladders Podcast episode number 97. This is a pretty special episode because I have never done this before, and this will be my very first monologue episode. You can call this section two cents, which we had a blog section on our website to begin with. So here I want to share some of my learning and thoughts on some of the interesting topics through my own media consumption and conversations because I think they're highly relevant and potentially valuable to this audience. My question today is ChatGTP making me stupid? Now, before we go into this very exciting topic, I want to announce an upcoming in-person event which is very, very rare for us because it's expensive to host an event. But we're planning a fantastic event on January 11th, 26th, yes, next year. And it will be Sunday, right before JP Morgan's. And this event is open to the public and we have limited space, so stay tuned for our official announcements through our newsletters and social media. This event primarily focuses on entrepreneurs and investors in the space of 3D technologies, in other words, for a pitch 3D audience. Now, back to the main programming. I have to come clean that we have published quite a few blogs on our website using, I would say 95% Chat GTP generated content. One example is the article for US Dental Schools at the frontier of 3D printing education. And weirdly, and perhaps not weirdly, I am not particularly proud of it. Yes, the content is good. It is valuable to some people based on the number of likes I saw on social media after we posted it. But me personally, I don't feel particularly happy doing it because I feel like it was kind of cheating. I haven't put enough efforts to earn the applause. So the first three people who can figure out which other blogs we have written primarily using ChatGTP on 3dheels.com can get a free ticket to our event in January. Okay, now let's talk about what I really want to talk about today, which is is Chat GTP making us dumb? You're probably already going to guess what my answer is, but the truth is this is a much more complicated question than it appears to be. And uh I think if you go with the flow of all the articles out there circulating in media right now, the conclusion is that ChatGTP is bad. No bueno, especially for students in the educational space. Students are not gonna learn as much or struggle as much mentally because of ChatGTP. And I want to talk about two articles I've seen to show some evidence of that, but I think this is a rather nuanced question and we should really think about slowly and carefully because we're at a juncture of technological revolution. And I hate to say this, revolutions typically will have some casualties. And many serious researchers are trying to figure out the good and evil of this very potent tool. That is, how can we use this new tool effectively to maximize our productivity without diminishing humanity and certainly not causing something unthinkable, like being ruled by an AI overlord. Okay, by the way, the latest scary scenario is not an AI overlord, but that AI does not even need humans at all. They don't need to rule over us because we are useless to them. So let's just pause and think about that a little bit. I have seen two fascinating publications I want to share here from academia. The first study is out of MIT Media Lab, explores the effect of using large language model or chat GTP for essay writing. It grouped students into three groups. The first one used Chat GTP to write essays. The second one, no tools, nothing. And there is a third group that used search engines. So each group completed three sessions under the same condition, under completed several different writing sessions under the same condition. And to gather the outcome of this uh of the effect of these tools, the researchers used EEG, electroencephalography, a brainwave monitoring device to assess cognitive load during essay writing sessions. They also used natural language processing to analyze the writings, both human and AI judges to score the essays. So here are the findings. You may be able to guess somewhat already. The no tool brain-only group exhibited the strongest, the most distributed network based on EEG, so highest mental loading, while the ChatGTP users display the weakest connectivity. The search engine group is in the middle, and this means the more tools you use, the less cognitive load. This makes complete sense. And here is a kicker. The self-reported ownership of essays was the lowest in the large language model group and the highest in the brain-only group. Large language model users also struggle to accurately quote their own work. Oh my god. That's exactly how I felt after I published those ChatGTP blogs on our website. I honestly could not recall all four schools that have named in that blog which dental schools focus on 3D printing. So, in this scenario, while large language model offers immediate convenience, this shows potential cognitive costs. And over four months in this study, the large language model users consistently underperform at neural, linguistic, and behavioral levels. So obviously, the conclusion of this article is that we need to exercise caution when use chat GTP for creative writing. The second publication came out of UPenn recently, focuses on solving math problems. And this study included over 1,000 students. The study showed that students using ChatGDP in the study were able to solve problems initially a lot better, about 48% over their peers. But then when tested without AI, their scores dropped 17% below their unassisted counterparts. This again tells us that a short-term productivity gain may have long-term performance cost. And if we as a species that aspires to progress through innovations and creativity, relying on tools like ChatGTP for near-term return on investment could be detrimental to our long-term goals. In other words, in the long run, you're not creating a more creative, robust mental labor group. So here I also want to introduce two interesting concepts I have learned recently. One is called cognitive offloading, and the other is cognitive miserliness. A bit of a tongue twist, but important concepts. So what is cognitive offloading? A cognitive load is the total mental effort used in a person's working memory at any given time. So for example, after the invention of calculators, we rarely need to do five-digit multiplication by hand. That is called cognitive offloading. After the availability of GPS and mobile phone, I rarely need to memorize route to my dog groomer, which is only three miles away, by the way. This is also a form of cognitive offloading. And Chat GDP is really one of the most powerful tools available for us to achieve cognitive offloading across numerous tasks in our daily lives. And I'm pretty sure you have tried out many of them. In fact, this is how AI increases productivity. We don't have to spend too much mental energy on mindless, repetitive tasks like organizing emails or putting things on our calendars. And here is the second, perhaps more important concept, which is cognitive miserliness. Okay, I have to look this word up, miserliness. It means excessive desire to save money. In other words, extreme stinginess. Mental miserliness is our tendency to do things that are less mentally taxing. So imagine you have to write an article like I did, with the help of Chat GTP, mental offloading or cognitive offloading, you could finish your task in no time and you really enjoyed that. And the next time you're gonna have to write an article, instead of struggling through this complex cognitive process, your brain just yearns for that chat GTP again, and over time become more reliant on that ChatGTP on a variety of tasks, that your ability to complete those tasks independently will atrophy. In fact, it is reported that some workers will use ChatGTP for work against their employers' wishes. It's almost like an addiction. So mental offloading or cognitive offloading can be good, but cognitive miserliness can be bad. So where is the perfect balance? I don't have a good answer for you, but I do want us all to be very aware of these because awareness is the first step. Avoiding a new technology is never the smartest move. Okay, I will not resort to learning riding a horse so that I can avoid driving, even though car accidents certainly kill. But I will fight hard to maintain the optionality to pick and choose which mental offloading I can cognitive or which cognitive offloading I can safely enjoy so that no other intelligence system can choose it for me. In other words, no thank you, AI overlord. Now I have to say a little bit about 3D printing. AI for 3D printing is a totally different ballgame, but it is uh is very different from LLM and it's still at its infancy. I have personally owned more than five printers of various quality and types in the past, and I have to say, I do not, do not love the experiences. I was mentally overloaded while tinkering with these machines, and for the moment I have given up on this. But 80% of people in this audience, perhaps, are not natural tinkers, or maybe I'm wrong. Let's just say 50% of the people in this audience are not tinkers. I can pretty comfortably say that. And that poses a problem for the industry. When your technology is too complicated to use, you are not going to scale. And now that we have some popular consumer-based printers like Bamboo Labs and Proof of 3D that are gaining tractions, they are still working in a group that has a predilection of being able to tinker. And the day 3D printing really becomes a mass consumer product is when we the machine is straightforward to use. Even a fool can use it. Like me. That I believe is where AI can really help us with some good cognitive offloading. All right, that's it for all today. See you next time. This podcast is for educational and informational purposes only. The views expressed do not constitute medical or financial advice. The technologies and procedures discussed may not be commercially available or suitable for every case. Always consult with a licensed professional.