If there’s one thing more hotly discussed than the UnAwards shortlists themselves, it’s surely Darren’s big reveal of the film shown on the day the gongs are handed out.
by Alex Waddington
While your host’s choice of attire - velvet jacket - is nailed on and never really in any doubt (no pressure Kings Heath dry cleaners, don’t mess it up!), he keeps the film a closely guarded secret.
But that doesn’t stop him being (nicely) pestered for clues, or having people try to guess it in the preceding weeks. In fact, I can remember having a few guesses myself in the car on the way back from Comms Unplugged in September. But he didn’t give anything away.
I’ve had the real privilege of getting to know D quite well over the last 15 months, since our paths crossed at CU22, and so I might like to think I have an insight into his cultural preferences, his outlook on life, and some of his favourite (and also least liked) things. But really, guesses as we headed back up the M6 were just that – guesses, informed by snippets of anecdotal data.
But wait. Because Darren is very open and refreshingly honest. He shares a lot of personal thoughts, experiences and reflections about what matters to him and what is important in life, on the brilliant comms2point0 blog and also in his eagerly read eMag in particular.
So, we do have a wealth of data, authentic intelligence and insight about Darren, going back over 10 years, which we can potentially correlate with film data to narrow down the possibilities.
And because the choice of the film is down to Darren and Darren alone, if we can learn more about him, we can probably increase our odds of working out what the film might be.
But my brain just isn’t powerful enough. I don’t have enough hours in the day to read through every one of Darren’s blogs, look at the previous films he’s shown, consult other relevant information about him in the public domain, and try to find common threads or themes between the various cultural reference points and data he’s given us.
But ChatGPT might.
I haven’t got an encyclopaedic knowledge of every film ever made, nor the plot lines, and I can’t possibly assess the entire contents of the IMDB and other such sites for suitability for the UnAwards event, the attendant assembled audience, the right length of the film, and so on.
But ChatGPT might be able to.
Or at least, a custom ChatGPT that allows us to link up specific information focused on Mr Caveney with the general, trained GPT model, and also with the facility to draw in live data from the internet for additional insights. The latest technology release and a paid £20 a month subscription has made this possible.
And so, in the spirit of fun (and fully approved by Darren) – but also to support some skills and knowledge development for me - the UnAwards 2023 Film Predictor, powered by OpenAI and ChatGPT, was born.
Data I used to create the custom GPT was as follows;
A scraped database of 150+ blogs tagged ‘by Darren Caveney’ posted on the comms2point0 blog since it started.
An interview Darren did with Vuelio
A few snippets like his DOB (found on Companies House - and no you’d never think it!), and where he lives and grew up.
A list of films he’s previously shown at the UnAwards
Additional snippets on his musical and film tastes from recent editions of his eMag.
Once I’d got this data into CaveneyGPT (I set the privacy so it doesn’t feed the general training model, even though there is nothing very sensitive in there), I published it to myself only, and started interrogating it.
First I asked it;
To examine all Darren’s blogs and pull out the films and music he mentions (which took all of 20 seconds)
To summarise the common themes he talks about in his blogs
Then I asked it;
To find common themes in the films shown previously.
To consider Darren’s professional background, his music, film and broader cultural influences, along with his previous choice of films – and give me a list of 10 possible films.
To also consider the fact it’s a professional Awards event, the likely celebratory and positive atmosphere and the audience attending
To filter out any 18s on the list with extreme sex or violence (out went Trainspotting at that point!) that might not be everyone’s bag
To assess the films on the list for IMDB rating and reasonable length
To consider there wasn’t an unlimited budget for screening rights.
By this stage, I was seeing many of the same films were maintaining their presence in the long list.
So finally, based on everything it had considered, I asked the GPT for a shortlist of 5, and then asked it to choose just one film that was most likely to be shown.
I printed out the result, put it in an envelope and posted it to Darren, to be opened just before the film is announced.
On Friday 1 December we’ll all find out the result of this experiment.
Whether it’s a triumph, a near miss, a broadly useful exercise, or a total fail, it’s been a fascinating process, and has given me myriad ideas for how we as communicators might harness the power of data and generative AI.
Today we are predicting the likely film choice of a legendary Brummie; tomorrow we might be helping hone the targeting and tactics for a major campaign that could save or change lives.
Maybe.
I’ll be back post-Awards to blog on what we might conclude from this experiment - and what learnings we can take into our everyday roles.
Alex Waddington is the Founder of Whetstone Communications and a data-in-communications advocate. You can say hello on Twitter at @TheAlexDubya
*Sign up for the comms2point0 eMag*
The comms2point0 eMag features exclusive new content, free give-aways, special offers, first dibs on new events and much, much more.
Sound good? Join over 3.4k other comms people who have subscribed. You can sign up to it right here
Image via Tullio Saba