DALL·E: A conceptual image of a diverse group of industry leaders engaged in a roundtable discussion about Artificial Intelligence in media analytics.

Industry leaders from AMEC member companies convened on April 22, 2024 in a roundtable conversation to discuss the integration of Artificial Intelligence (AI) in media analytics. 

The event was hosted by AMEC Board Directors Simon Gebauer (Observer Media) and Maya Koleva (Commetric), with coordination from AMEC EU Chapter Co-chair Sophia Karakeva (DataScouting). The roundtable special guests were Alicja Bors (Ruepoint), Nesin Veli (Identrics), Michal Hrones (Newton Media) and Stavros Vologiannidis (DataScouting). The open format invited contributions from the audience and open questions.

Leveraging AI in Media Measurement

The roundtable began with Simon Gebauer seeking insights into the current use of AI within the industry. This prompted discussions about AI’s role in enhancing functions like tagging, sentiment analysis, and summarizing extensive media content to provide digestible insights for clients.

Alicja Bors from Ruepoint highlighted the tailored approach their team adopts in developing AI solutions for each client, emphasizing the importance of continuous interaction and feedback. “It’s crucial to maintain a strong, ongoing connection with the client and continuously refine AI tools based on their feedback,” she explained.

Diverse Client Reactions and Strategic Adaptations

The discussion also touched on the mixed reception of AI technologies among clients. Alicja noted variability in acceptance, with some eagerly adopting AI tools, while others remain reserved, waiting for clearer regulatory guidance. This underscores the necessity for transparency in integrating and communicating AI’s role in media products.

Nesin Veli from Identrics discussed the proactive interest clients have in generative AI and large language models (LLMs), highlighting their potential impact at various levels from personal to company-wide implementations. “A lack of enthusiasm on the part of clients is something that we have never complained of in the PR and communication industry”, Nesin jokingly said. According to him, clients are very engaged and keen to utilize the latest AI technologies, despite the potential risks.

Ethical and Legal Considerations

A substantial part of the conversation was dedicated to the ethical and legal challenges of employing AI in media analytics. The experts discussed the legal landscape surrounding AI-generated content, emphasizing the need for careful navigation to balance innovation with compliance.

Nesin Veli advocated for using on-premise solutions: “Everything is vacuumed off the internet and everything is done on premises, where we can control the input and the output. And really these smaller custom built models – the results so far show that they can do the specific tasks well without breaking any legal and ethical codes. These models are oftentimes more interpretable than the large models that you can use.”

Stavros Vologiannidis from DataScouting highlighted the crucial first steps in AI integration, focusing on data ownership and the appropriate platforms to leverage AI effectively. “The first step in adopting AI is ensuring you have control over your data and the right infrastructure to support AI technologies,” Stavros noted. He also discussed the broader implications of AI, from traditional applications like image and speech recognition to more advanced uses involving large language models.

Technical Challenges and Future Prospects

Michal Hrones of Newton Media brought up the technical challenges related to integrating AI within existing data systems, stressing the importance of developing sustainable AI models capable of managing vast datasets without quality compromise.

Stavros Vologiannidis added to the conversation by underscoring the importance of scalable AI solutions. “Scaling AI applications requires significant resources, not just in terms of technology but also financial investment. Combining smaller language models with traditional techniques can provide a balanced approach to leveraging AI in media analytics,” he explained.

In the conclusion of the roundtable, Simon Gebauer introduced a final, intriguing question from the audience, challenging the panelists to choose just one Large Language Model (LLM) for all their media analytics work. The responses highlighted diverse preferences and strategic considerations unique to each panelist’s organizational needs and experiences. Alicja Bors diplomatically declined to specify a single model, emphasizing the variety of tools at her disposal. Nesin Veli advocated for custom, in-house models tailored to specific use cases, underlining the importance of data ownership. Stavros Vologiannidis echoed this sentiment, expressing satisfaction with both Mistral and Lama models due to their alignment with DataScouting’s data sovereignty principles. Michal Hrones also favored bespoke solutions developed internally to leverage proprietary data effectively. Maya Koleva, while a proponent of OpenAI’s models for their robust capabilities, stressed the necessity of adapting the choice of LLM to the specific data security and client requirements of each project, underscoring a commitment to continuous experimentation and innovation in the use of AI technologies in media analytics.

Lookout for more conversation on AI in just a few weeks’ time – see the AMEC Global Summit 2024 programme – and register now for in-person or virtual attendance!

Listen to the full roundtable in the AMEC Member area here.

The article is produced by a CustomGPT based on a full transcription of the roundtable conversation.