To cut through the often-frenzied headlines, we sat down with Andrew Bruce Smith, Chair of CIPR’s Artificial Intelligence in PR group and digital communication consultant, to uncover more about what ChatGPT means for the industry.
Why has ChatGPT garnered so much attention since its launch, when the likes of Jasper AI have been commercially available for a long time? What’s new now?
This is a question I had on my mind only the other night, asking what has happened and what’s transforming the conversation. Large language models (LLM) – like GPT – have been around for a while. Microsoft has made big moves so far with ChatGPT (it’s an investor in parent company OpenAI). Facebook has its own LLM but has not yet done anything publicly. Interestingly, Google has recently announced plans to launch Bard, its own competitor to ChatGPT.
But I look back to 2014, when I started running training workshops for both the CIPR and PRCA. I remember saying to agencies back then to look at AI and machine learning-driven writing tools, such as Wordsmith, which at the time were writing thousands of automatically generated articles for the Associated Press. It wasn’t a real human journalist; this was a machine writing these articles. But of course, many of the stories were very template driven. So, things like financial reporting lent itself to Wordsmith: ‘Company X reports revenue of y, that’s X percent up or down. Compared with period Z.’
What ChatGPT has done is effectively given access to anyone who knows how to use a chat window. If you could write a few lines of API code, you could have done the same thing already with GPT. But how many PR professionals can code? ChatGPT simply provides the chat interface to anyone who wants to utilise it, and that’s the real kind of ‘Aha’ moment when you effectively democratise access to these powerful capabilities.
So, is it a friend for foe?
It could be either. It’s up to the industry to decide. Ignoring it is technically an option – there are clearly some people who’ve decided it’s a fad or a gimmick. These are likely people who have tried it once using a bland input and receiving a bland output. I don’t think they’re in the majority.
In many ways, the only way we’ll learn how to harness it is to simply use the tool, try it out, have some fun, and experiment with it. You’ll find out by doing what works and what doesn’t, and you’ll understand what it’s capable of and what its limitations are.
Understanding what it can and can’t do will hold PR professionals in much better stead than fearing it or throwing the towel in, calling it game over. Those that don’t are the ones that think the machines are going to rule, and go and live as a hermit on an island and unplug from the Internet.
ChatGPT can easily – and should be – a huge, powerful complement to what PR people do anyway, and should want to continue doing.
Google recently announced an alternative to Microsoft-backed OpenAI’s ChatGPT. Is it right to feel threatened?
First, it’s important to remember that most people write about Google from the outside. But it’s undeniable that Google themselves – along with pretty much the entire tech industry – has been taken aback by ChatGPT and the incredible combination of uptake of use and interest in it. Google isn’t exactly sitting there twiddling its thumbs on AI; they’re clearly one of the frontrunners and have been one of the leading researchers and exponents of it.
Dennis Habib, CEO of DeepMind, said in an interview a few weeks ago that the company is working on a similar AI tool named Sparrow. He admitted that he could have released it months ago, years ago, but chose not to because of his views on the implications of the technology. It’s all very well to say ‘hey, put it out there and see what happens’ versus ‘let’s be a bit more cautious, because we’ve got no idea what some of these unintended consequences might be.’
But I think it’s fair to say that Google feels slightly threatened. That’s why they’ve since come out and announced they’re going to release multiple new AI-related products this year, including Bard, a ChatGPT-style interface that will be incorporate into search. There’s a whole raft of things that Google is looking at. But it’s not just them. Look at Facebook. It’s interesting that Facebook’s Head of AI came out and said that ChatGPT is not as revolutionary as you think, and that they’ve got a similar product that just hasn’t been put out there yet. It strikes me as sour grapes because ChatGPT did get out the door and gain the headlines. And now they’re wishing that they were out a bit earlier.
Which other day-to-day duties, aside from copy and content production could ChatGPT – or any other AI tool – disrupt?
ChatGPT was launched on 30 November. And a week after, with almost zero fanfare, Google quietly slipped out a little (free) add-on for Google Sheets. And it might be fair to think ‘big deal, what’s a spreadsheet add-on going to do to the world?’
But despite its simplicity, what it basically does is allows you to now run machine learning yourself on your own data. That’s huge. There’s been a lot of talk over the last decade about big data and AI. About how that can be applied to machine learning to big datasets and detect patterns and insights that no human being is capable of finding out, which is great. But I think what’s tended to happen is that the ability to do this has really only been available to organisations with big datasets and huge computing resources. The fact that anyone with a computer and an internet connection in the browser can now go into Google Sheets, integrate the add-on in a few seconds, and go, right, here’s my own data. That’s game changing.
So, I was trying to think of some of the practical PR use cases. For example, let’s say you’ve got a spreadsheet full of media coverage data, including estimated coverage views. Let’s imagine some cells have missing figures, and the client is pushing for answers – what would you do?
You can get machine learning to plug in the gaps, based on the data you do have. But it gives you those numbers with a level of statistical confidence. If it came back and said, look, you’ve got 50% of your data missing, but we plugged the gap in 45% of it with a 98% confidence level. That’s just one example.
That’s a classic example of something that is now freely available to anyone that wants to use it. And it’s out the door at the same time as ChatGPT. And no one seems to have paid that much attention to it.
There is not an hour that goes by without some new innovation or use case. GPT is now available as a spreadsheet function. So, for example, let’s say you wanted to generate 15 alternative headlines for this piece that you’ll write. What you could do is tell the model that you’ve got to write 14 different articles around 14 different subjects. At the push of a button, the integration could deliver you the alternatives that could make your piece stand out. That GPT integration is now also available for Google Docs.
The mind bends at the level of sophistication you can get to. The ability to do this didn’t exist a month ago, and now it’s like ‘whoa, what’s going to happen in the next four weeks, the next four months, the next four years?’ I think it’s time to embrace it, as it’ll only help you do even more impactful work.
Check out the full Great Minds series here.