Early experiences with Generative AI

To set a context for the conversation, our media experts shared their early experiences with Gen AI. There was general agreement that Gen AI marks a potentially huge step forward in many areas and is likely to prove highly disruptive in the short term as its application is explored. Most companies are unwilling to stand still and wait for the wave to hit them, rather they are keen to understand and be prepared for the likely magnitude of change.

Pro Sieben, the German media group, has been involved in AI for some time as an area of investigation for their data strategy teams, engineers and specific AI product teams. The focus has been on activities that can be automated, but also on applications for content creation and program planning and scheduling.The keywords that apply to Gen AI at the moment are ‘experimentation’ and ‘exploration’. Some media execs are happily throwing themselves in at ‘the deep end’ while others are more cautious, but all have at least some experience and early insights. There is real innovation underway that could impact operations from content to media sales, administration and human capital development."

"I think we're in the early days. Experimentation, exploration. There is considerable interest – we ran an internal data & tech conference recently and had multiple sessions on AI. It is definitely the hot topic of 2023!”

Margo Swadley, BBC Studios

Our media leaders flagged up a number of important issues:

The need for clarity: Gen AI is more than just machine learning

As highlighted in our introduction, there is a lot of potential confusion about where ‘traditional’ machine learning ends and Gen AI begins. There can be a danger in activities having the label ‘AI’ slapped on them as a rebrand, much as many statisticians found themselves rebranded as ‘Data Scientists’ a decade ago. The situation is not helped by AI vendors themselves, with the vendor landscape described as the ‘wild west’ in the rush to market new tools. As Craig MacDonald of McKinsey relates, conversations with clients often need to start with agreeing on what AI actually is:

“I like the separation between traditional machine learning and how Generative AI is adding to that because I have found that with a lot of the work that I do, there's a 10 minute conversation about “what are we talking about here? Let's just call machine learning Generative AI now and rebrand everything?” You know, once you get through that conversation, it comes down to what's the impact on content generation, which is purely a generative AI application.”

Craig MacDonald, McKinsey

However, it’s important to note that the advent of Gen AI does not somehow supplant wider applications of Artificial Intelligence, which has always been an umbrella term for many technologies that use deep learning techniques.

“Although we're focusing on Generative AI today, I'd like to highlight that AI is an umbrella term for many use cases. While different definitions vary in details, machine learning and optimization are commonly seen as important branches of AI.”

Gufeng Zhou, Meta

The need for guidelines and principles

Media organizations should be able to benefit from their employees' enthusiasm to investigate and experiment with Gen AI applications, but need to avoid the potential pitfalls in the process.

As a result, companies are working on internal guidelines for their staff to help them avoid any risks involved in the use of open-source tools. In the UK, the IPA - the media agencies’ trade body - is already looking to establish frameworks that will give its members guidance. It is also involved in initiatives with ISBA and the Advertising Association, seeking to establish guidelines as members look to them to help navigate the new terrain by understanding best practices and the legal implications of using Gen AI.

Ultimately it will be vital that media organizations are able to empower their brightest minds, but also mitigate the considerable risks involved in any new technology that disrupts and challenges traditional legal frameworks.