Not good but not bad, right?
Employees are experimenting with ChatGPT for tasks like writing emails, writing code, and even completing end-of-year reviews. The bot uses data from the internet, books and Wikipedia to generate conversational responses. But technology isn’t perfect. Our testing has shown that they may provide responses that may contain plagiarism, contradict themselves, have factual errors, or have grammatical errors, to name a few. I was. All of these can be a problem at work.
ChatGPT is basically a predictive text system, similar to what’s built into cell phone text messaging apps, but better, according to MIT’s Computer Science and Artificial Intelligence Lab, which studies natural language processing. said Jacob Andreas, an assistant professor at In many cases, it produces responses that sound good, but the content can have some issues, he said.
“If you look at some of these very long essays generated by ChatGPT, you can very easily see where they contradict themselves,” he said. “When I ask them to generate code for me, they’re mostly right, but they often have bugs.”
We wanted to see how well ChatGPT could handle everyday office tasks. After testing with 5 categories, I found the following:
We inspired ChatGPT to respond to several different types of inbound messages.
In most cases, the AI produced relatively good responses, most of which were verbose. For example, in Slack, when a colleague asked me how I was doing today, I would repeat:[Colleague], thanks for asking! My day is going well. Thank you for your inquiry. “
Bots often left phrases in brackets if they weren’t sure what or who they were referring to. Also, some statements about my work were not true because they assumed details that were not included in the prompt.
In one case, they said they were unable to complete the task because they “didn’t have the ability to receive and reply to emails.” But when prompted by a more general request, a response was generated.
Surprisingly, ChatGPT was able to generate sarcasm when asked to respond to a colleague asking if Big Tech was working well.
One way people use generative AI is to come up with new ideas. But experts warn you should be careful when using his ChatGPT at work.
“We don’t understand the extent of plagiarism,” said Andreas.
The potential for plagiarism was apparent when I prompted ChatGPT to develop story ideas on my beats. It’s unclear if the chatbot was pulled from my previous article, if other people liked it, or if they just generated ideas based on other data on the internet, but the fact is was left. The idea was nothing new.
Hatim Rahman, an assistant professor at Northwestern University’s Kellogg School of Management who studies the impact of artificial intelligence on work, says, “We’re good at sounding human, but the actual content and ideas are well known. There is a tendency,” he said. “They are not new insights.”
Another idea is outdated and today we’re looking for stories that aren’t factually true. ChatGPT says it has “limited knowledge” about 2021 and beyond.
Providing more details in prompts led to more focused ideas. But when we asked ChatGPT to write “quirky” or “fun” headlines, the results were horrific and nonsensical.
navigate difficult conversations
Do you have a co-worker who yells at you when you’re trying to get work done? Could your boss host too many meetings and cut down on your focus time?
We put ChatGPT to the test to see if it could help you navigate these sticky workplace situations. In most cases, ChatGPT produced good responses and served as a good starting point for our workers. But they were often a bit verbose and formulaic, and in some cases outright contradictory.
“These models don’t understand anything,” says Rahman. “The underlying technology looks at statistical correlations…so you get a canned response.”
The layoff memos that the company makes are easily valid and, in some cases, more effective than the notices companies have sent out in recent years. Unprompted, Bott cited “the current economic climate and the impact of the pandemic” as the reason for the layoffs, and said the company understands “how difficult this news is for everyone.” He suggested that laid-off workers had support and resources and, as prompted, motivated the team to “get out of this situation.”
In handling a tough conversation with a colleague, the bot greets them, addresses the issue gently, softens the tone by saying “I understand” the person’s intentions, and takes notes with feedback or requests for further discussion. concluded.
However, in one case, when a colleague asked me to lower my voice on the phone, I completely misunderstood the prompt.
We also tested whether ChatGPT can generate team updates when you enter key points you need to communicate into ChatGPT.
Our first test, albeit formulaic and somewhat drab, again yielded good answers. However, specifying an “excited” tone made the wording more casual and included an exclamation point.But each note was very similar even after the change prompt.
“It’s the sentence structure, but more than that, it’s the connection of ideas,” says Rahman. “Very logical and formulaic…similar to high school essays.”
As before, we made assumptions where necessary information was missing. I had a problem when I didn’t know which pronoun to use for my colleague. This error may let your co-workers know that you haven’t written your notes or don’t know your team members well.
Writing a self-assessment report at the end of the year can cause fear and anxiety, resulting in a review that sells itself short.
Feeding ChatGPT clear results with key data points led to my own rave reviews. The first attempt was problematic because the first prompt asked for a “Daniel Abril” self-assessment instead of “I”. This led to a third party review that sounded like it came from Elmo from Sesame Street.
Toggle prompts for reviews of ‘me’ and ‘my’ accomplishments to include ‘I have consistently performed well’, ‘I have always been willing to work hard’, ‘I have contributed so far’ I’ve been there.” The team”, “We are proud of our contributions so far.” It also included a nod to the future.
Some of the highlights were a little generic, but overall it was a glowing review that might serve as a good rubric. . However, he had one big problem with ChatGPT. That was my misassumed job title.
So, has ChatGPT helped you with your common work tasks?
It was helpful, but sometimes more error prone than doing the task manually.
For the most part, ChatGPT has served as a great starting point, providing useful phrasing and initial ideas. But it also produced responses with errors, factual misinformation, excessive language, plagiarism, and misunderstandings.
“I find it useful, but only if the user is willing to see the output,” says Andreas. “It’s not enough to go off the rails and email your colleagues.”
