If you need to update a published session report or add attachments (images, files, links), send a note to osullivan@dvb.org.
AI Agents in a DVB-I world
Host: Rafael Bermúdez, 3Cat Reporter: Rafael Bermúdez, 3Cat Session goal: We are heading towards a world full of agents, and we need to see how we can create an appealing ecosystem for users, manufacturers, and broadcasters.
Summary of discussion:
For this session, we created a role-playing scenario in which participants were assigned different roles. There were three roles: TV Assistant Agent, Broadcaster Assistant Agent, and DVB-I. The aim of this exercise was to visualize how communication occurs between these elements.
When a user asks, “Which program am I watching now?” to a TV Assistant Agent, we observe that the assistant can resolve the query using DVB-I information. Similarly, a Broadcaster Assistant Agent can answer the same question independently. In this case, there is no need for the two assistants to communicate with each other, as each actor can resolve the request using its own tools: the TV Assistant natively, and the Broadcaster Assistant via an HbbTV app and a companion device (since a microphone is not available within HbbTV).
On the other hand, when a user asks, “I want to watch a Bruce Willis film” to a TV Assistant Agent, the assistant attempts to respond using the information available to it, including DVB-I endpoints. However, a key issue arises: not all broadcaster catalogs are accessible through DVB-I endpoints. As a result, some content remains hidden from the TV Assistant Agent.
Even if full catalog access were available through DVB-I, there would still be technical challenges in keeping all broadcaster catalogs continuously updated within an intermediary layer managed by the manufacturer to support search functionality.
The conclusion is that, to provide a successful content search & discovery user experience, where all broadcaster content catalogs are accessible through TV Assistant Agents, further solutions are required.
Note for all participants: If something is missing or seems like an opinion rather than a fact, feel free to reach out.
Next steps:
Ping to TM-I working group to ask if this is something they want to take