Last week, we started experimenting a bit with OpenAI’s ChatGPT, which ignited a LOT of discussion amongst our team about its potential pros and cons for our industry and others, as well as how it can be used for good (and evil). Read on for our initial thoughts!
Landis: What are your impressions of ChatGPT’s responses to our prompts about AI and PR last week?
Polly Ikonen: So, there’s writing a blog post, and then there’s writing a blog post anyone would want to read. The post generated by ChatGPT here – granted, without any stylistic direction – reminds me very much of my 10th grade composition class. Mrs. Ledbetter, decked out in head-to-toe polyester (this was waaay back), drilled into us the structure of a five-paragraph essay: introduction with thesis, topic sentences, supporting points, all the way down to the conclusion that reiterates the thesis. Rinse and repeat. A terrific foundational exercise for a 15-year-old. A much less compelling read for a 30-year PR veteran. (See what I did there? I used some incomplete sentences – ChatGPT’s were all complete. Yawn.) ChatGPT’s blog lacks specifics, anecdotes, cultural references, stylistic oomph. It’s not wrong, but it’s generic – and consequently of limited utility.
The one response that gave me pause was in response to the final question Diana posed: Do you have any stats or data to support this blog post? I found ChatGPT’s response both disappointing and exemplary:
ChatGPT: I’m sorry, but I don’t have access to any statistics or data to support the blog post. However, there is a growing body of research on the use of AI in the public relations industry that you may find helpful. Some studies have found that AI can be effective in automating certain tasks, such as content creation and media outreach, and can help PR professionals to better understand the effectiveness of their efforts through data analysis. It is worth noting that the use of AI in PR is still relatively new, so more research is needed to fully understand its impact.
It’s the perfect non-response response, the kind we train our clients to be able to deliver when they don’t know or don’t want to share the actual answer. It’s disappointing that ChatGPT couldn’t point to specific data, given that it “knows” there’s a growing body of research. Yet it’s an interview dodge worthy of a real-life PR pro.
Landis: Are there upsides to ChatGPT?
Diana Haven: Oh, absolutely. I look at ChatGPT as a work buddy or assistant that is here to help all hours of the day, on any topic imaginable, in any language, instantly. For example, if you are suffering from writer’s block, just enter a prompt related to your project, and you’ll be off and running. I wouldn’t recommend copying and pasting its results verbatim for several reasons (AI detectors can tell it’s written by AI, for one), but with some editing, refining, additional research and adding your personal finesse, it helps save loads of time. You can do so much more, too, such as menu planning based on what’s in your fridge, your particular diet or special occasion; writing cover letters in your voice for specific jobs tailored to your resume; creating draft legal contracts; writing code for a new app; translating whole documents into another language; or even writing poems about your cat in Shakespearean English. It’s not going away, and future versions will only be more powerful. It’s already generating a whole new class of websites, start-ups and innovations. I’m personally very excited about and slightly terrified of what’s to come.
Landis: What are some of the disadvantages of using ChatGPT?
Gretchen Krueger: What I’ve noticed is the lack of creative content. There is no personalization or anecdotes in the final product. It lacks the human or empathetic touch. The content output is only as good as the content input, so there are likely many gaps and biases, as well as misleading and incomplete data when using ChatGPT.
Also, what about jobs? Will ChatGPT replace specific tasks usually completed by people, or will it replace jobs held by people altogether? Not everyone has access to technology and internet connections, so ChatGPT may not be available or reliable for all in all situations.
And just for fun, I went to the source to ask what it thinks are the downsides to using ChatGPT:
There are a few downsides to using ChatGPT or other large language models, such as:
- Computational resources: Training a model like ChatGPT requires a significant amount of computational resources, such as powerful GPUs.
- Memory requirements: Once trained, ChatGPT and other large language models have a high memory footprint, which can be an issue when deploying the model in a production setting.
- Lack of interpretability: Because large language models like ChatGPT are based on neural networks, it can be difficult to understand how the model is making its predictions.
- Bias in data: the pre-trained model is trained by a large text data which can be biased in a sense that it reflects societal bias, stereotypes and more. This is taken into account when using the models in certain sensitive areas.
- Requirement of fine-tuning: Pre-trained models like ChatGPT can produce generic response which might not be task-specific, so fine-tuning on task-specific data is often required to achieve good performance.
Use ChatGPT responsibly!
Landis: What are your impressions of ChatGPT’s impact on social media thus far?
Brianne Murphy Miller: ChatGPT might make waves in content, but does it have the snark to match, say, Ryan Reynolds…
Apparently, yes, as Reynolds used ChatGPT to write a new ad which he read on his social media channels for wireless provider Mint Mobile. Of course, the fact that he asked the bot to include a curse word could have given ChatGPT the clues it needed to be snarky. You can view the video here.
What does this mean for social media? Well, with the right queries, you might be able to replicate your social tone and voice, but the intrinsic authenticity, which is a requirement in the social world, means you’ll have to make sure to review after you automate. My prediction? The many new clickbait accounts out there using ChatGPT might make those little blue check marks even more valuable. Real person? Check!
Landis: What is ChatGPT’s effect on journalism?
Robin Carr: When it comes to journalism, ChatGPT could potentially be dangerous, especially in this era of “fake news.” In an article recently published by the Business Insider (“ChatGPT and the Future of Trust”), senior business reporter, Samantha Delouya, sounded an alarm: “I asked ChatGPT to do my work and write an Insider article for me. It quickly generated an alarmingly convincing article filled with misinformation.”
Yikes. Two things to note:
- Mainstream media is crying foul on the assault on truth. Fact-checking seems to be a thing of the past, and rarely is there accountability among fringe media.
- How many journalists could ChatGPT replace?
Dr. Lili Mou, Chair at Amii (Alberta Machine Intelligence Institute) and an expert in AI, says, “For traditional media and particularly factual correctness, I think journalists will still have to check copy grammatically, logically, and factually.”
And as far as replacing journalists, Mou decided to ask ChatGPT:
ChatGPT in journalism is likely to change the nature of the work rather than replace human journalists entirely. While the technology may be able to automate certain tasks, such as generating rough drafts of articles, it is unlikely to be able to replace the creativity, insight, and human judgement that are essential to journalism.
You said it, ChatGPT. Let’s hope it’s right.
Landis: What is ChatGPT’s current and potential impact on education?
Andie Davis: With the emergence of AI tools like ChatGPT, many educators are worried that their students will turn to AI to do their assignments for them. However, some argue that students and educators should embrace AI technology in the classroom, as it will only become more prevalent and robust over time.
According to EdWeek, the average teacher works 54 hours per week, but spend less than half that time teaching students. One plus of ChatGPT is that it is FAST, offering a quick and easy way for educators to create lesson plans, quizzes and much more. Some are also creating assignments using the AI tool to have students fact check if what it is producing is accurate. I think this is a great idea considering Landis recently tested ChatGPT to see how accurate the information was, and we found at least one factual error in the copy the program provided.
On the flipside, using ChatGPT for assignments can hurt students in the long run because homework provides an opportunity for critical thinking and practicing concepts learned in the classroom. If a student consistently relies on tools like ChatGPT to do their work and think for them, what have they really learned?
I can see both sides, but ultimately, we can’t avoid AI in the classroom forever. I think it will eventually become the “norm” as the technology is fine-tuned and integrated into everyday life, but we must be cautious to ensure that it is providing a net benefit for students and educators alike.
Looking ahead, I think the question really should be: how can we use tools like ChatGPT in the classroom to enhance students’ education and improve the educator experience?
Landis: What are the fun and creative uses for ChatGPT?
Makenzi Jordan Rodriguez: Every writer is familiar with the frustrating standstill caused by writer’s block. Before ChatGPT, what was the solution? Time: take a break, read some inspiring literature, or run your draft by a trusted colleague. With ChatGPT, prompts can be quickly generated into relevant ideas and content to provide fresh inspiration.
In business terms, time is money, yet ChatGPT’s rapid iteration and content proposals also supplement the creative process by generating ideas for nearly any task. Hosting a last-minute Trivia Tuesday party and need a quick list of questions on a variety of subjects, say, popular 1970s clothing items? Here’s (a few of many Q&A) what ChatGPT came up with:
- What type of shorts were extremely short and popular among women in the 1970s?
Answer: Hot pants
- What was a popular style of sleeveless top that tied around the neck?
Answer: Halter top
- What was a popular decoration on clothing and accessories such as vest, jackets, purses and boots?
Answer: Fringe
ChatGPT vaguely describes its role as a creative tool for “a wide range of fun and entertaining applications, and the most fun use would depend on the specific needs of the user.” That’s exactly right – if you can think it, ChatGPT can help you do it, from planning your dog’s birthday party to drafting the next screenplay.
To generate a good idea, you need a lot of ideas, and ChatGPT has the capacity to combine and recombine more possibilities than our minds can. This tool is not to be dismissed but rather artfully wielded, and yet the risk lies with creatives: those who choose not to use ChatGPT technology may fall behind their more cost-and time-efficient contemporaries.
All things considered, the technology of the future is here, and we must prepare to cautiously embrace AI as it continues to incorporate into our personal, professional, social, and educational spheres, and more.