From Assistant to Asset: The Promise and Perils of ChatGPT as a Learning and Professional Tool
I didn’t write the title for this article. Instead, I typed in ChatGPT ‘s search bar “give me a catchy title about the implications of ChatGPT in education and the workforce.” ChatGPT is a large language model developed by OpenAI that I’m sure many of you are familiar with. The AI tool was released by nonprofit research company OpenAI on November 30, 2022, and is only gaining more traction through its ability to curate mostly accurate responses to virtually any question. Its ability to cater so well to the user experience has got Google running with its tail between its legs, threatening Google’s search business and opening the world’s eyes to getting answers with AI chatbots.
With the release of GPT-4, there’s this AI-hype train that everyone is trying to get on. There’s been an impressive number of AI tools that people are creating, and keeping up with it all is the biggest challenge(a problem an AI tool could also solve):
Several articles and tools surrounding ChatGPT tend to focus on the technology and taking advantage of potential as an “office assistant.” However, what we’re failing to consider is the undeniable impact on the future generations, whose world is becoming increasingly digitized. What does this mean for the millions of students and children who now live in a world where information is so readily available? How will this impact their eventual entrance to the workforce? And what about the millions of teachers and professors whose carefully-crafted lesson plans are starting to lose merit thanks to ChatGPT?
In an effort to answer my questions, I have to admit something else: I didn’t come up with the idea of using ChatGPT to create the title of this article. That credit goes to NYU Professor Pascal Wallisch. As a professor of psychology at NYU for eight years and professor of data science since 2020, his perspective of ChatGPT is beyond valuable, not only for teachers desperately trying to work around the AI tool, but also for students looking to effectively integrate ChatGPT into their lives. To really dive deep into what ChatGPT does for this world, I asked him for his thoughts on the AI tool at the beginning of the semester, as well as how he plans to approach it as a professor. Pascal also gave advice for data science students trying to navigate their careers. Below are some of the highlights from our hour-long conversation:
First Impressions and Causes for Strong Retention
“What impressed people was because it was so inaccessible before that, only the experts knew how far these things had already come.”
“I saw it immediately, because of my Twitter, social media, as well as through other data science people. I saw it the day it came out and tried it the day it came out because I already had an OpenAI account from the DALL·E OpenAI art bot… as it so happens, I was teaching a machine learning class in the CS department in spring, so this was actually a very relevant topic. So I actually had to cover DALL·E on an emergency basis because it was kind of disrupting the art space at the time in the spring.”
Professors, especially those in data science or AI/ML, address new AI tools as they are released. Even so, there’s always this thrill of exploring these tools’ functionalities that professor Wallisch reflects on: “It was mind blowing because that night I shared it with my lab… and someone in the lab said, let’s ask it to write some code. Just right there. And, I remember everyone going “whoa.” Because first of all, it was correct, the code was correct, but there was also a link on the upper edge. You just could copy and paste that into your console, then it would be running code. That moment, everyone realized this is gonna change everything.”
Despite the initial “wow” that the rest of the world shared with professor Wallisch , he says there’s more to why ChatGPT seems to have a continued grasp on the world’s attention. “I think the reason is twofold. One is you could say GPT3, which is what ChatGPT is based on, has been around for like a year now, right? So a lot of people who are in this space were like, yes, we have known this for a year now, but I think what made ChatGPT so transformational was the accessibility. You don’t have to be a machine learning expert to do this. You just can type it in. You don’t have to have any knowledge of how this works.”
ChatGPT has become another one of several prime examples as to why UI/UX design is so crucial to a technology’s success. This leads to professor Wallisch’s second point, being that “what impressed people was because it was so inaccessible before that, only the experts knew how far these things had already come. So I think a regular person was just very surprised that it can do all kinds of things that are actually useful to regular people.”
Funny enough, this tends to be the case with most technologies. An example that Wallisch brings up is cars. There are nearly 250 million licensed drivers in the U.S. [1], but how many of them understand the inner workings of a car? Even computers: “In the early seventies, you basically had to be a computer scientist to operate a computer.” I’m a CS major and I couldn’t tell you every part that makes up the laptop I’m currently using to type up this article(oops). But really, professor Wallisch says that what makes the ChatGPT situation unique is “this jump between what people thought bots can do and what bots can actually do.”
Creating Equal Opportunities for All Students
Conversations surrounding ChatGPT in education tend to go south because of ChatGPT’s ability to write well-reasoned 5-paragraph essays at a high school and even college first-year level on virtually any prompt. This is why an understandable primary concern with ChatGPT for all teachers is cheating.
While Wallisch acknowledges its shortcomings, he is quick to identify the good that is also created through ChatGPT, especially for students facing accessibility barriers. “Let’s say someone is ESL (English as a Second Language), and by the way, I’m ESL. I’m very sympathetic to that. And they have good ideas, but they just can’t express them because they’re not fluent in the language. Well that probably just went away. What if I wanna write in Japanese or something like that? I can do that. It can translate into other languages too. Any language really.” In fact, Wallisch suggests that people stick to learning one language, not just linguistically but in computing as well. “Why would you wanna be mediocre in five languages if you can be really, really great in one and then have just a bot do everything else?”
When we think about students who tend to experience learning disadvantages, we’re not just talking about ESL. What about students with ADHD, who may experience difficulties starting an assignment (and perhaps this also applies to students without ADHD)? “Is it ethically so bad if the bot a) keeps you on track and b) gets you started?” Perhaps we need to turn our attention towards ChatGPT’s tool-like nature as a crutch for students. In a way, ChatGPT can “level the playing field” and allow students with all kinds of mental makeups to excel.
ChatGPT Exposing Educational Institutions
“How much of our education system was just geared at systematically creating sophisticated sounding bs?”
Of course, the previous point doesn’t take away from the fact that ChatGPT potentially creates an environment where students don’t have to learn. The signature compare and contrast book A and book B essay question that I’m sure all of you remember from English class is out of the picture. Wallisch even emphasizes that one of the best features of ChatGPT is its ability to summarize. Need a page summary of 1984? ChatGPT could do that for you. If you want a summary in 3 pages, it could also write it for you.
It’s no surprise the level of social backlash that ChatGPT sees because of this. Wallisch says it’s similar to the art community’s reaction to DALL·E, another AI bot developed by OpenAI that creates AI-generated images based on a user’s input. Similarly to how society chooses certain things to frown on (e.g., being naked in public), the art community now “frowns” upon the creation of AI-generated art.
So, what are some proposed solutions in an educational setting? First and foremost: a ban. NYC schools have already banned ChatGPT from school devices and networks [2]. However, Wallisch argues that attempts to ban such a useful tool cannot be properly implemented. “Do your students have a cell phone? You know, students have smart watches. There’s no way. How are you gonna ban this? And if it’s not ChatGPT, there’s a lot of other bots that are now in the works. And this is gonna be integrated into the search engine, this is gonna be integrated into the world…” The last part of this line cracks me up. We had this conversation 2 months ago, and Microsoft just recently launched its ChatGPT integration with Bing.
Perhaps ChatGPT is revealing something somewhat more sinister about the education system. If the things we’re taught or meant to learn in school are so easily replicable, then maybe we’re not learning the right things. Think about it — how much has the education system changed since it was first made compulsory in the U.S.? This was back in the mid to late 1800s, when states were slowly making education a requirement for all U.S. citizens post-industrial revolution. While it is true that curricula have changed, especially as you go higher in education, the structure of educational institutions persists.
Wallisch compares the way ChatGPT is trained to how parrots are trained to “have conversations” with humans. “A parrot might be saying hello, and the first time you talk to a parrot, it’s very eerie actually, because the parrot doesn’t have to understand what it’s saying.” In a way, ChatGPT operates similarly. The response that ChatGPT outputs is the one it has the highest confidence in after performing complex pattern detection of similar responses to similar questions. It doesn’t actually know the meaning of what it’s saying, but as Professor Wallisch says, “*we* give it meaning.” Doesn’t this just make you wonder whether “most students are parrots and they just write stuff that they think the professor wants to hear?” An even scarier question Wallisch asks is “How much of our education system was just geared at systematically creating sophisticated sounding bs? Because the bot is good at creating that, and if they are indistinguishable, you can draw your own conclusions.”
Adjusting to a New Normal for Professors and Students
Maybe the whole “our education system isn’t actually teaching us anything” debate is highly controversial, but it really raises the question of how teachers and professors should teach from now on. It’s not fair to make exams or quizzes so hard that only the top 5% of students do well, but this also means giving students take home assignments must be done knowing that they have ChatGPT at their fingertips. “I used to have a fully open book, use everything you want, final exam. And the idea was, well, you know, the employer will not lock you into a room without any resources.”
Wallisch shares some of the steps he’s taken to teach a more ChatGPT-friendly semester this spring. “My new final exam will be, I mean, I don’t like to do this, but fully locked down and in-person. The idea is to simulate an interview or emergency situation, where there is literally no time to confer with a bot. We will check IDs, no phones, no computers, no nothing. Maybe you can bring any notes you want, but if we see an electronic device, we just assume you’re cheating with the bot. But anything before that final exam, we assume that for those things you take home, you will use the bot. So we expect more.”
Even though Wallisch feels his new way of administrating the final might be perceived as unrealistic, preparing students for entering the workforce is a priority. “Do you think employers are gonna pay you for anything that a bot can do for free and without needing breaks or vacations? Obviously not, so you have to show a little bit more.” There’s this harsh reality that comes with any new technology that doesn’t even have to do with jobs outside of the field that may get replaced. Even for data science students, the stakes are higher.
But what specifically in regards to data science can ChatGPT replace? Well, with proper integration, there’s no need to write your own SQL queries. In fact, much of the data gathering, formatting, and cleaning can be done by ChatGPT. Let’s not forget about its ability to write clean and accurate code snippets in virtually any language! In fact, if you need instructions on how to go about training a neural network or model using a certain algorithm, ChatGPT can thoughtfully generate code blocks to run. There are just limitless forms of automation and code generation that ChatGPT is able to accomplish.
In case you’re a data or computer science student, and you’re noticing your palms start to sweat, don’t fear (too much). Professor Wallisch says that in his own lectures, he focuses on what data scientists are really needed for– judgment and decision making. “Given a dataset, let’s say management has a question about their business, can we help them make decisions? And that is something that the bot cannot do yet.” The reality is you can tell ChatGPT what to do, but only after deciding on what kind of model to use, which coding stack an application you’re building will run on, or performing some level of critical thinking, which just became more valuable.
Other skills that are becoming more valuable include a combination of manual and social skills. “For a bartender, you have to have cognitive skills like mixology, social skills to talk to people at the bar, and manual skills to make the actual drink. A plumber, for example, would have a motor component, a social component, and a cognitive component.” However, Wallisch says that careers in accounting, legal writing, or careers that mostly rely on bs-ing at scale might be at risk in the long term.
Final Thoughts
Overall, ChatGPT is an extremely promising tool, and if you’ve made it to the end of this article, then you’re probably open-minded like myself and Wallisch. Throughout our entire conversation, I noticed that an idea of “mastery” often came up. We don’t need to kind of, sort of know several languages anymore, we only need to be fluent in one, and ChatGPT will translate the rest for us. This also applies to programming languages — if you’re really good at Python, then writing something in C++ shouldn’t be too difficult with the help of ChatGPT. Even for data and computer science majors entering industry, there’s an emphasis on having a strong understanding of the field to the point where you can make impactful technical decisions for companies that ChatGPT can’t do. Lazy-developing can be replaced, but the creativity, logic, and communication that go into all software projects cannot. Nonetheless, with every technology comes some level of criticism.
What do you think about ChatGPT? Is it leading us to our doom or sparking one of the most influential tech booms of the century?
References
[1] Hedges Company (2023). https://hedgescompany.com/blog/2018/10/number-of-licensed-drivers-usa/#:~:text=How%20many%20licensed%20drivers%20are,from%20238.6%20million%20in%202022
[2] Rosenblatt, K. (2023, Jan 5), “ChatGPT banned from New York City public schools’ devices and networks.” NBC News. https://www.nbcnews.com/tech/tech-news/new-york-city-public-schools-ban-chatgpt-devices-networks-rcna64446
Bhogal, R. (Interviewer). (2023, January 13), Pascal Wallisch.
