Fiona Field tested out the artificial intelligence chatbot to see how effective it was with a brief and to see if it lives up to all the hype
Probably the most interesting development since the dawn of Google was the launch of ChatGPT last November. The roll out was soon followed by a host of competitors. Recently, I have been dabbling with its pilot tool to test its capabilities and understand its limitations. The results are both exciting and depressing in equal measure. What is ChatGPT and will it change what we do?
ChatGPT stands for ‘Chat Generated Pre-Trained Transformer’ and is an artificial intelligence-powered language model developed by Open AI. Microsoft have reportedly invested over $10 billion in the hope that it will upend Google’s dominance. From limited early experiences, one thing is abundantly clear, the new technology will tear up the internet as we know it.
What sets ChatGPT apart is the flow of conversation which feels human-like. The more specific you are in defining the task, the better the output. Even on first use, the results are remarkedly different than how you would engage with a search engine. For example, I tasked ChatGPT with writing an opening paragraph on what it can do, using an informative style of no more than 50 words.
Briefed on electric car ad: Fiona Field posed an advertising challenge to Open AI’s ChatGPT chatbot. The Ask: Please write me a headline ad to sell an electric car to customers in Ireland using humour. The Answer: “Get your Irish jig on with our electric cars – the only ride that’s as green as your shamrock shake!” Pictured are members of the Q102 road crew with one of the station’s Nissan Leaf electric cars on the north quay of the River Liffey.
“Introducing ChatGPT – the latest breakthrough in AI language models. Built on the cutting-edge transformer architecture, it’s trained on massive amounts of text data and can generate natural-sounding responses to text-based prompts. Whether you’re looking for a chatbot to power your customer service or a virtual assistant to assist with your daily tasks, ChatGPT has the smarts to deliver.”
Not a bad effort, even if it did ignore my word count. It’s easy to see why universities will have to navigate this latest tech in the fight to counter plagiarism. In less than six months we know that ChatGPT can pass both law and medical exams in a host of prestigious universities (albeit with varying results), create lyrics for songs in the style of famous artists and compose scripts for plays.
Yet not everyone is smitten.
Nick Cave responded to a ChatGPT song developed in his style with three words: “This song sucks!” Cave went on to say that AI song writing is a “grotesque mockery of what it is to be human”. While he may have been spitting fire at using tech to reduce the craft of song writing to a piece of code, the same could have been said about the birth of the electric guitar back in 1936.
While the uses of Open AI are vast from in how it can deliver customer experiences, or write blogs to improve SEO rankings, perhaps its biggest challenge is the reliance of historic information and algorithms that might be biased. Like any piece of new tech there are shortcomings and limitations but the potential for how this may evolve could help transform the industry.
Much of the written word across the web is not of the Harry Potter variety, whether that’s a manual for your dishwasher or a description of a restaurant. Yet despite the possibilities for AI, we should remember the mistakes of the past. Human oversight and brand safety is paramount. You may remember Microsoft’s Tay back in 2016, which was a Twitter chatbot mired in controversy.
It was all due to the inappropriate and racist tweets that Tay generated; all of which happened in less than 24 hours. While AI is still in its infancy, how the creative industry will transform will be an interesting journey. Offering a glass half full perspective, the fragmentation of media has been a major challenge for the creative industry which sparked a long line of disruption.
A quarter of a century ago, one TV ad would have lasted a decade and there was no real need for fluent devices, owing mostly to the dominance of the media of the day. Today, it is vastly more difficult for advertisers and agencies, both commercially and logistically, to produce enough content to satisfy the myriad of channels and consumer appetite across the board, particularly on social media.
Considering the billions of dollars and the race to monetise what’s next for big tech, I’d be optimistic that like most inventions AI will evolve and get better over time to serve many uses for the creative industry. Will it be a friend or foe? I will leave you to ponder the Chat GPT-style generated ad that I plugged into which summed up both the surprising potential and scary limitations.
Fiona Field is managing director of OMD