By Richard Bagnall, Global co-managing partner and CEO of Europe and the Americas at CARMA.
Generative AI has come out swinging. Who remembers the computer-generated newsreader from the Noughties Ananova? Now it’s back, bigger and better than its rudimentary predecessor - including a Kuwaiti News outlet debuting an AI-generated news presenter.
This isn’t the only application of generative AI across media and communications. CNET is embroiled in AI controversy with suggestions it has published articles written using AI tools. Snapchat has gifted all its users a GPT-powered AI chatbot. WIRED now has a generative AI policy on its website detailing what it may do (use AI to generate story ideas) and what it would never (publish regular stories generated by AI).
2023 has already shifted the needle on how we read, tell and share information. So what does the evolution of generative AI mean for communications, and what pitfalls must we desperately avoid?
The power of generative AI
What is generative AI? Put simply by McKinsey, it ‘describes algorithms (such as ChatGPT) that can create new content, including audio, code, images, text, simulations, and videos’ and falls under the umbrella of machine learning. As such, its applications are almost endless. As WIRED’s policy notes, generative AI can help creatives spark ideas or suggest optimal headlines or social media copy. Bing and Google are using generative AI in their search functions. Canva has built a new set of AI tools, and Damien Hirst’s recent AI-generated project resulted in over $20 million in revenue across nine days.
Generative AI is bringing new art and creativity into the world. But one of its big advantages is taking on less imaginative tasks to allow humans more time to focus on work of greater importance. For example, PR and marketing agencies must often produce large amounts of written content. Generative AI could support here by outlining and structuring first drafts that professionals can quickly polish into final pieces. And earlier this year, a BuzzFeed internal memo revealed plans to use AI to create content, specifically around its well-known personality quizzes.
The pitfalls of generative AI
However, the huge potential of generative AI means it must be treated with an equally large degree of caution. Choosing to ignore the many challenges in its use could be catastrophic.
One key problem is the data generative AI is trained on. As AI-generated content rises, so do copyright concerns: in April, the FT revealed that Universal Music Group told streaming platforms to stop AI from scraping melodies and lyrics from their copyrighted songs. But an even bigger foundational concern revolves around the bias in AI. It’s just wrong to think AI is infallible - even though a piece of code is ‘emotionless’, it’s written by biassed humans and trained on data selections humans provide - some trained robots have shown traits of becoming racist and sexist.
Generative AI can make significant mistakes with breathtaking confidence. A news agency experimented with using ChatGPT to write an article detailing a crime. The AI got pretty much everything about the incident wrong and even fabricated a quote from the victim’s children. In a world already wrestling with misinformation in the age of social media, the ethical implications of this are enormous. Not only is there the risk of unintentional misinformation, but generative AI also opens the door to deliberate misuse.
‘Deepfakes’ (using AI to imitate someone through video or audio) have been around for a while. Platforms like TikTok are popularising the use of AI to mimic celebrities and even political figures, from Joe Biden to Joe Rogan. The consequences of AI-generated misinformation may be reputational or financial for organisations, but politically, they could even be life or death.
Generative AI and the Future of Communications
The power and pitfalls of AI mean legislation and regulation are desperately needed. This is something government bodies are working on: the European Commission started drafting the AI Act almost two years ago, and it’s now entering the trialogue stage. It’s set to classify AI tools to a level of risk and demand high levels of transparency from those using ‘high risk’ tools. The concern is governments face a steep mountain to climb in terms of legislating as fast as AI is innovating.
But what about worries around potential job losses that always accompany news of tech automation? It’s hard not to think of BuzzFeed’s AI memo in light of the recent closure of BuzzFeed News and planned redundancies across the rest of the company. Yet BuzzFeed is just one instance of the mass layoffs sweeping across media and journalism, from Vice to Paper. The glaring mistakes of AI demonstrate that creative work cannot be wholly, simply handed over to a computer - AI needs rigorous checking, guiding and finessing by human hands. The human power to provide critical thinking, context and reasoning will still be in demand. For better and for worse, generative AI is a tool that can extend human capabilities and experiences, not necessarily shrink them.
Mark Twain famously said only two things are needed in life: ignorance and confidence. AI can have both in spades. It’s up to the humans who use it to harness the latter and control the former as we enter this new age of generative AI.