Research Spotlight: The rise of ChatGPT and the age of artificial intelligence

Jules White, associate professor of computer science

When one of his Ph.D. students first suggested that Jules White, associate professor of computer science, check out ChatGPT—the artificial intelligence platform that can do everything from write original poetry to generate sophisticated computer code in seconds—White was dismissive at first. But once he investigated further, White knew that this technology would shape the future, not just for his own field, but for numerous industries and applications, drawing comparisons to the dawn of the internet itself and mobile computing.

Here, White provides an overview of ChatGPT’s capabilities and offers insights about what impact its adoption may have on students, computer scientists and the future of technology as we know it. 

For those who don’t know what it is—or what it’s capable of—how would you describe ChatGPT?
Rather than describe what it is, which won’t do it justice, I’m going to give an example of what it can do. With ChatGPT, you can have the following complex conversations with your computer:

Me: “I am allergic to gluten, I don’t want to spend more than $100/wk on groceries, and I like spicy food. I don’t want to eat more than 2000 calories per day. Please provide me with a meal plan and list of recipes that will fit my needs.”

ChatGPT: “Here’s a meal plan and list of recipes that could fit your needs:

Breakfast

Scrambled eggs with sliced avocado and salsa

Gluten-free oatmeal with sliced banana, cinnamon, and honey

Snack

Carrot sticks with hummus

Roasted chickpeas with paprika and cumin

Lunch

Gluten-free wrap with turkey, lettuce, tomato, avocado, and hot sauce

Spicy lentil soup with gluten-free crackers

Snack

Apple slices with almond butter

Gluten-free rice cakes with hummus

Dinner

Spicy chicken and vegetable stir-fry with gluten-free rice

Grilled salmon with gluten-free quinoa and roasted broccoli

Snack/Dessert

Gluten-free dark chocolate or fruit salad

Me: “Excellent, but I am going to have a hard time keeping track of my groceries for this. Can you write a simple software application in Python that I can enter my initial list of groceries into and then let you know each time I cook one of the recipes? It should let me know which ingredients to buy at the end of the week.”

ChatGPT: “Sure, I can help you write a Python application for this. Here’s a simple solution that you can modify as per your needs:

# Initialize the groceries dictionary

groceries = {‘Eggs’: 12, ‘Avocado’: 2, ‘Salsa’: 1, ‘Gluten-free oats’: 1, ‘Banana’: 3, ‘Cinnamon’: 1, ‘Honey’: 1, ‘Carrots’: 1, ‘Hummus’: 1, ‘Chickpeas’: 1, ‘Paprika’: 1, ‘Cumin’: 1, ‘Gluten-free wrap’: 1, ‘Turkey slices’: 6, ‘Lettuce’: 1, ‘Tomato’: 2, ‘Hot sauce’: 1, ‘Red lentils’: 1, ‘Onion’: 1, ‘Garlic’: 2, ‘Carrots’: 2, ‘Celery’: 2, ‘Gluten-free crackers’: 1, ‘Salmon fillets’: 4, ‘Quinoa’: 1, ‘Broccoli’: 1, ‘Dark chocolate’: 1}

# Initialize a dictionary to keep track of which ingredients have been used

used_ingredients = {}

# Print the initial grocery list

print(“Initial Grocery List:”)

print(“=====================”)

for item, quantity in groceries.items():

print(item + “: ” + str(quantity))

Did you recognize ChatGPT’s potential significance immediately? Or was there a single moment or experience that made you realize this could be a game-changer?
Up until now, there have been two inventions that I immediately knew would have a sweeping impact on our society: 1) the Internet and 2) the iPhone. The first time I saw each of these things, I immediately knew that they were going to completely change our day to day lives.

ChatGPT is the third invention that will have that kind of impact—possibly more. ChatGPT is an artificial intelligence application that can answer questions and generate answers that appear to contain very sophisticated reasoning. Every domain that works with documents, text, and code is going to change. Also, any domain that relies on computers will change, because ChatGPT can help you interact with a computer.

The moment I started writing code with it, I knew ChatGPT was a game-changer. I could ask it to change the code it produced to meet high-level design goals regarding structure, modularity, and maintainability that were never possible before. Its ability to take high-level feedback and improve its output is what really sets it apart in my mind.

And its potential impact is extremely broad. For example, my wife is a Speech Language Pathologist who administers early childhood language assessments. One of the first questions that really changed my understanding of its capabilities was when I asked ChatGPT to: “Write a report for an early childhood speech language assessment for a three year old with autism named Jim. Include the specific assessments that you used and the scores. Provide examples of interactions with the child regarding communication.”

It generated a report that my wife said was exactly like what she would produce. It listed the correct assessments, like the PLS-5 and ADOS-2, and used a very similar format to what she would use. It can’t interact with the child and do an assessment, but it certainly has potential to help Speech Language Pathologists like my wife write reports.

What kinds of conversations have you had with your students about ChatGPT?
Someone told me, “AI isn’t going to take your job, someone using AI is going to take your job.” This type of tool is going to be a huge productivity enhancer for the people that really understand how to use it effectively. Vanderbilt needs to graduate students that are expert ChatGPT users and have all the critical thinking skills and knowledge that we already teach them.

I tell my students to absolutely use it, but to cite where they use it. Next fall, I am teaching a course on how to rethink the way we write software now that we have ChatGPT.

Writing an essay for a literature class with ChatGPT is one thing. What about writing code, like your students are doing — how well does it handle that?
With the right training and understanding, ChatGPT can make someone 10-times more productive at coding. The key is to really understand how to talk to ChatGPT to produce high-quality code that can be maintained. The gamechanger for me is that I can have a conversation with ChatGPT about the code and it can rewrite the code to achieve certain desirable properties better than many software developers I know.

ChatGPT is still driven by the creativity and insights of the user. A bad software developer isn’t going to get much better with ChatGPT. But a great software developer is going to get immensely more productive. The productivity boost will allow them to explore more in code and try out more ideas than they code have before.

There has been great demand for computer science graduates over the past 10, 20, 30 years—do you see ChatGPT replacing any of these jobs?
Not yet. These are still tools that help a human, but don’t operate fully autonomously. The tools also aren’t good enough yet to allow a non-programmer to build anything other than fairly simple software applications. Real, large-scale software applications require architectural insight and creative structuring that these tools can’t quite handle yet.

We may create even greater demand for computer scientists. I predict that we are going to produce a lot of software using these tools over the next five years. A lot of that software is going to have bugs, design issues, and security flaws that you can’t just ask ChatGPT to fix. We will need computer scientists to do these critical tasks that require a ton of reasoning. We are also going to need computer scientists to build on top of ChatGPT and create all of the new industry-specific applications with it.

How do you see industries making use of ChatGPT in the months and years ahead?It is really hard to answer this question. It is a bit like being in 1984 and trying to predict how the Internet is going to change industry. I would say that for most industries, if you aren’t using ChatGPT or whatever equivalents get released by competitors in the next few years, you are going to get replaced by the the company or person that is using ChatGPT. This type of tool has such wide-ranging impact that you aren’t going to be able to compete without it.

Thinking about the wider spectrum of AI and machine learning, is ChatGPT the holy grail, or just the tip of the iceberg. What other major developments do you think we’ll see in the next 2-5 years?
It’s likely the tip of the iceberg. It is probably the most exciting thing that I have seen in my career. I view it as an enabler of my creativity. I feel like an artist that has been painting on cave walls and is suddenly given an iPad, Apple Pencil, and an inkjet printer. There is still beauty in cave paintings, and they won’t go away, but the new technology is going to open up all kinds of new possibilities for my creativity to explore.