Actually, AI IS Coming for Your Job, Especially if You Are in Tech

Human + AI > AI

By Troy Lowry

The next time you have a problem with your computer, iPhone, or tablet, almost any problem, try solving it with ChatGPT first. Simple or complex, it is a great help.

For instance, recently, a neighbor wanted a desktop shortcut on their Windows machine to make it easier to connect into her work Zoom calls. It had been maybe a decade since I created such links, so I knew that it was easy but didn’t remember the exact process. I asked ChatGPT, and it gave me the process. Instead of fumbling around for ten or fifteen minutes trying to remember how it worked, within two minutes I looked like a genius.

Some of the big advantages ChatGPT has over in-person helpdesk support:

  1. Always available: 24/7 support that answers instantly.
  2. Non-judgmental: You don’t have to worry about it judging you for asking it the “dumb” questions. It’s always happy to respond.
  3. Consistency: I’ve been impressed at how ChatGPT gives the same responses time and again.
  4. In-depth knowledge: I have fed ChatGPT very obscure error reports, and it has helped me find the cause.

That said, there are some areas that ChatGPT falls far short in, so be careful!

  1. It can suggest some rather extreme “fixes.”  If it suggests you delete apps or programs, think twice. If it suggests you reset your device to factory configuration, don’t!
  2. It does not always have the right answer. If your issue is common, it will almost certainly have a good answer. But if your issue is unique or rare, it probably will not.1
  3. It can’t take control of your machine and make the fix itself the way many human help desks do.
  4. It is programed to give you an answer when sometimes the best thing is to ask more questions. For instance, if you don’t specify what sort of device you are on it will likely give you fixes across a number of different operating systems.

As good as ChatGPT is at helping with computer problems, it’s even better at coding. In an earlier blog post I talked about fine-tuning ChatGPT to have it write more like me to fool the ChatGPT detectors.2 To make that happen I had to write code to call OpenAI’s fine tuning APIs. I quickly encountered some odd errors and did what I usually do, search the internet for answers.

That’s when I had my epiphany; I decided to show the error to ChatGPT and have it tell me what to do. To my great delight, it immediately told me that one of the packages I was using would not work and why.3 I then asked it to write the rest of the code, which it promptly did. It did not work flawlessly, but I put the error back into ChatGPT, and then it did work flawlessly, saving me hours of work.

For coding, a specialized Large Language Model, called GitHub Copilot, was designed specifically to help coders. A coder simply opens a file, describes in comments what behavior they would like to see from the code, and the AI automatically generates it. Copilot overtime learns your coding style, which is great to keep code consistent.

My experience is that Copilot is a lot like ChatGPT, extremely helpful but not 100% reliable. Just as having ChatGPT write an email for you would be a mistake unless you carefully review the output, the same applies to having it or Copilot write code without carefully checking the results. Often, only minor adjustments are needed, but those adjustments are crucial.

The pace at which artificial intelligence is evolving is nothing short of staggering, especially its potential to reshape the landscape of technology jobs. While it's true that AI platforms like ChatGPT and GitHub Copilot are not infallible, their ability to provide instant, round-the-clock assistance is revolutionizing the way we approach problem-solving and coding. These tools are not just supplementary aids; they're game-changers in the industry. Yet, it's important to underline that human expertise, combined with AI, produces results that are far superior to what AI can achieve alone. The symbiosis between human intuition and AI's computational power creates a dynamic that is greater than the sum of its parts.

As AI continues to advance, it's not crazy to predict that it will have a transformative impact on tech roles, automating certain tasks and optimizing others. While some might view this as a threat to job security, it's better framed as an opportunity for professionals to elevate their roles, focusing more on strategic, creative endeavors that machines can't replicate. In the era where AI is a reality, not a distant future concept, learning to coexist and collaborate with AI isn't just an option; it's a necessity. This partnership of human and machine promises to redefine what's possible in technology, bringing us into a future filled with unprecedented possibilities.

  1. Oh, it will still give you answers. It might even give you a dozen things to try. It is programmed to answer. But the answers won’t be useful and could even be harmful.
  2. And it worked beautifully. While those ChatGPT detectors didn’t work so well to begin with, training ChatGPT to write in my style by feeding it hundreds of examples of my prose fools the detectors pretty much every time.
  3. To be more technical, I was trying to run everything using curl commands from PowerShell and was running into issues. ChatGPT told me the issue was that PowerShell’s Invoke-WebRequest was aliased to curl and wouldn’t work for what I needed. Instead of removing the alias, which might cause other issues, or calling curl.exe directly, I decided to just use Python instead.

Troy Lowry

Senior Vice President of Technology Products, Chief Information Officer, and Chief Information Security Officer

Troy Lowry is senior vice president of technology products, chief information officer, and chief information security officer at the Law School Admission Council. He earned his BA from Northeastern University and his MBA from New York University.