Generative AI: a new phase of innovation
‘Transformational’, ‘augmenting’, ‘magical’ – these are the words that were peppered throughout WPP’s roundtable on generative AI. We asked what this technology means for the creative industries
The application of advanced technology to human insight to deliver the best solutions to specific tasks – and do so rapidly – is what generative AI delivers. It is not a single system or algorithm. It is a combination of data, models and predictions that can deliver new creative outputs: images, video, text, audio and code. It is already used by WPP across numerous campaigns, largely to augment and accelerate the creative process – but it is always used in synch with human talent and insight.
“Generative AI is an evolution of people working with machines to create content,” said WPP’s Stephan Pretorius. “For us, the magic occurs when you combine human insight – and cultural insight – with this ability to generate content with machines.” It is the spark that occurs when humans and machines interact. “This is WPP’s role. We apply these technologies, combine them with insight, and help our clients grow.”
What does AI offer that traditional man/machine techniques do not? “Speed,” said Microsoft’s Matt Groshong. “It’s the ability to accelerate and react to whatever’s happening in the world today, and to deliver something exceptionally creative. It can also take you down a path that you had not yet thought of.”
WPP’s Di Mayze concurred: “Generative AI rewards the user for being intellectually curious and articulate. It has the ability to surprise you.” That is why creatives from across WPP have leant in and adapted to using these tools quickly and enthusiastically. It facilitates idea exploration.
But it also gets ideas closer to execution faster. “Generative AI allows you to go from imagination to representation fast, and it allows more time spent on imagination and less on execution,” said Pretorius. If the development of concepts – storyboards – is fast, creative teams can choose how best to put them into production quicker too.
The historical data conundrum
As machines are clearly trained on historical data, how do you make sure they are trained on representative data that you can trust? WPP’s Vicky Brown responded: “From a legal perspective, there has to be transparency. You have to understand how vendors have built their tools and systems, and how they have trained their tools. Then, when you use those tools, you must think carefully about your inputs. Of course, you must then sense-check the outputs too; and you need to understand vendors’ terms and conditions so that you know the use limits of your outputs.”
At WPP, ongoing, robust training has been in place since 2019 to ensure users of these tools think about the use of personal data, data privacy laws and confidentiality from the get-go. “Privacy laws require users to think carefully about the information that goes into the system. But these are not complex legal theories at play. It is about making sure that you input sensibly. And then, think carefully again about the output for intellectual property issues, ethical issues, taste and decency,” said Brown.
Clearly, legal and regulatory frameworks are playing catch-up with the technology itself, largely as a consequence of the speed with which generative AI is developing. “That is why we have to police ourselves,” said Brown. “We have created our own set of guard-rails; we have our own policies for working with creative AI; and we have a security and privacy charter. We have to get the regulatory issues right on behalf of clients; and we also have to understand clients’ risk appetites,” said Brown.
“Let’s not forget that the generative AI tools that are making the headlines are trained on all the content in the world,” Pretorius reminded us. “There’s a big question about content curation – it’s a challenge. Increasingly you will see managed data sets for training purposes, particularly in relation to copyrighted material.”
A brand’s perspective on responsibility
While there is clearly responsibility on the creative industries, their technology partners and their professional advisors to act both responsibly around inputs and thoughtfully around outputs, ASOS’s Papinder Dosanjh says brands have a responsibility to monitor their systems too – to make sure the right alerts and checks are in place.
Dosanjh leads the data science and machine learning teams at ASOS. This team builds the machine learning systems that power the online fashion retailer’s digital experience. “We often monitor for drift in the data, which will change the outcome of the models,” she said. “The way in which our models are performing day in, day out, minute by minute, is something that we check forensically. And we work with our legal and cyber security teams to make sure that the decisions we're making are the right decisions for our customers.”
But there will always be conversations between WPP and its clients about risk and how to mitigate it. That is a natural part of the WPP/client relationship. But it is more than that: shared insights in relation to risk will benefit the entire business landscape and, ultimately, across consumers too. Microsoft shares this view.
Groshong said: “We have not kept our processes around responsible AI inside Microsoft. We have published them. And we have worked with WPP to help train its people. But this was not just about training WPP people, it was also for our benefit to enable us to learn and listen. The basis for our responsible AI programme is principles, practices, tools and governance – all four elements have to be in place.”
Dosanjh concurred: “Transparency and education in our business is very important. We have to be able to make decisions explainable at every level. And complex systems and models make that very hard to do; but we make sure that, even with a deep learning system, we can do that.”
Advanced tech requires talent
Much of the progress we are seeing in the commercial application of generative AI comes about through relationships and trust. But there is also a need for the type of talent that can grasp this new normal. “You can't work with these technologies at enterprise level if you don't have people who can understand these technologies at a very sophisticated level,” said Pretorius.
“We’ve invested not only in training our people at WPP, but we’ve also acquired a company with primary AI research capabilities – Satalia. These are key capabilities for companies to develop if they are going to work in these areas. This is not to minimise the skillsets of creative people – you just need these additional skills too.”
It is the thirst for talent that is driving the ecosystem approach – no single company can build the future of AI. And thankfully, this collaborative approach may well ease concerns around AI-generated code – hard to spot and probably best kept for lower risk functions. How much AI-generated code will be open source is yet to be seen – but there has been a trend in this direction.
Where on the curve of adoption are we?
“This phenomenon is really capturing people’s imagination. There is a sense of wonder,” said Pretorius. Mayze concurred: “If you think about humans being enhanced by AI, and then add creative thinking, we are going to continue to be surprised and delighted by generative AI. Using this technology will enable us to see things totally differently.”
Groshong added: “AI is becoming fundamental to the tools, platforms and capability that we are releasing. Our mission to help people achieve more. AI is leading that ambition.”
What is more, over time, generative AI will not be limited to enhancing the creative process, it will also increasingly enhance business processes and utility. Ultimately, it will be about empowerment.