AI and ChatGPT – friend or foe?

Generative AI has gone into overdrive, and the only certainty is that the pace of change will increase rather than slow down over the next few months – but what does that actually mean for those of us who are having to deal with these challenges?

This is far from an anti-AI rant (quite the opposite in fact). It’s going to change our lives and be a valuable toolbox to help us. But whilst the number and functionality of AI tools are growing exponentially, that doesn’t mean it’s time to turn over all your marketing and comms work to ChatGPT. It’s not about to give you real strategic advice or a ‘hold the page’ campaign concept.  

The communications and marketing profession absolutely should be using generative AI tools where they help us do our jobs more effectively and in a secure and ethical way – at DTW we’ve already been using a number of AI tools for years to assist with things like the transcription of audio content, managing website enquiries and as part of the production process for video and animation content, but that’s a very different use of the technology. 

Most people’s skills as ‘AI prompt engineers’ (future job title alert!) are in their infancy, and while everyone will learn by experimenting, that doesn’t mean now is the time to put all your eggs in the AI basket.

Part of the value we at DTW will provide to clients moving forward will be insight and guidance to help you through the AI minefield, but for now, here are six practical thoughts and tips:

Don’t give out confidential information

Rule number one – don’t use ChatGPT or an equivalent to create a written summary of your new unpublished report or confidential financial data. It isn’t worth the risk of giving the AI system access to training data about your organisation or project which it could potentially share with other people.

Refining and repurposing content can be valuable

Using large language models like ChatGPT as a starting point can be valuable in helping to refine or repurpose content, but it should be as a supporting role to a human, and never with confidential or sensitive information! Tools like Microsoft Copilot will soon be rolling out within the Office applications we use every day making this process seamless and an integrated part of our working practices.

Ethical disclosure

Transparency is important – no one should pass off AI generated content to your boss or client as something you’ve done. It won’t end well. Any agency should be very transparent with clients about how they are using AI, particularly when it comes to generative AI tools like ChatGPT. Anyone tempted to proudly pass off an ‘AI creation’ as ‘their own work’ is not someone you want doing your strategic comms for a whole range of reasons (not least if they can’t write more engaging copy than ChatGPT then they’re in the wrong role), but it can have a role – just like an online thesaurus sometimes does the trick when you’re looking for inspiration. 

Check all generative AI content carefully for accuracy

Just like humans, AI isn’t perfect and definitely makes mistakes – in fact, it’s been known to make things up that are entirely fabricated. Whether that’s factual inaccuracies brought about by using training data from pre-2021, incorrect assumptions based on what it thinks is a good idea, or simply the AI model choosing what it thinks is the best string of words based on an algorithm, some of them can be really silly and just downright wrong. It’s good at joining dots together, just not necessarily in the right order.

Consider copyright concerns 

This can apply to words, images, video or concepts. You won’t know where the AI tool you are using has taken the solution from that it has given you. There could be huge legal, financial and reputational consequences. If you are looking to create something ‘original’, then AI isn’t a magic bullet to solve your challenge. Tools such as Adobe’s Firefly AI image generator attempt to deal with issues of copyright by using licensed training data from the Adobe Stock library, but equally suffer from this by not having the same breadth of source material as alternatives such as OpenAI’s Dall-E which shows in the results.

Remember the value of nuance, originality and critical thinking 

A lot of AI generated content looks and sounds very similar. Ask ChatGPT to create a ‘marketing plan’ for three very different organisations and you don’t get anything that reflects the different challenges, contexts and opportunities they face. Part of the value that a good agency or comms team should provide is in spotting the challenges or opportunities that aren’t obvious. 

Want to dive in and learn more but not sure where to start – check out www.futuretools.io to see what might be useful for you (but set aside some time – there were 1,635 listed as of late May 2023), as well as the CIPR’s recent Artificial Intelligence in Public Relations report as a starting point for further reading.

If you want to know more about what we’re doing behind the scenes to experiment, trial and test on your behalf then get in touch.