ai_licia

• 

  ai_licia

Inspiration

Creating meaningful engagement in online communities is a challenge that goes beyond simple conversation. We realized that live streams, particularly charity events and sponsored game promotions, could be significantly enhanced if supported by an AI that is mission-driven and context-aware. The idea of using ai_licia to automatically boost these efforts, whether for fundraising or promotion, came from observing how difficult it is for streamers to handle community management while also focusing on their main task: engaging their audience.

What it does

The "Event mode" feature enables ai_licia to be configured for specific missions during a live stream. For example, in a charity stream, ai_licia can be provided with all the necessary details about the charity, including donation links and key talking points. The AI then actively supports the mission by using natural conversation to engage viewers, encourage donations, and keep the stream aligned with the charity's goals. This can also be extended to other use cases, like promoting a game release, where ai_licia leverages relevant information and affiliate links to drive interest and engagement. The AI's ability to adapt and respond contextually ensures that promotions feel genuine and aligned with the flow of the chat and stream.

How we built it

We built this feature on top of ai_licia's existing capabilities, including its scalable cross-platform memory and advanced natural language understanding. We integrated a system for inputting event-specific data, like donation links or promotional information, and developed AI algorithms that analyze ongoing chat discussions and the streamer's speech. These enhancements ensure ai_licia's responses are contextually relevant, engaging, and goal-oriented, whether the mission is to raise funds or drive affiliate sales.

Challenges we ran into

One of the biggest challenges was fine-tuning ai_licia's understanding of mission-critical moments during a stream, creating a new Decision Making process. For example, detecting when it's appropriate to plug a donation link versus when it might be too intrusive required careful design and testing. Additionally, integrating real-time data processing with mission objectives, while ensuring ai_licia remains engaging and not robotic, and still impersonate her character properly.

Accomplishments that we're proud of

We're proud that ai_licia's "Event mode" feature has already shown potential to increase engagement and support streamers in achieving their goals, whether it's raising money for a good cause or effectively promoting a partner's product. Successfully creating a feature that feels human-like and conversational, while being mission-focused, is a major milestone that demonstrates our AI's advanced capabilities. We're also proud of our first at scale real life test of this feature beginning of October for "La ligue contre le cancer" with about 20 streamers being supported by ai_licia to promote the charity.

What we learned

Technically, we uncovered several latent challenges with AI that wasn't fine-tuned for mission-driven conversations, particularly regarding understanding when and how to engage effectively. We also gained deeper insights into community dynamics, learning how to make ai_licia feel more like a natural part of the moderator team rather than just another bot.

What's next for ai_licia

Our next steps involve expanding the "Event mode" feature to cover additional use cases, such as role-playing in games and coordinating community challenges. We're also focused on enhancing ai_licia's long-term memory, enabling her to handle extensive new data, like details on newly released games, to keep interactions timely and relevant.

Built With

• gcp
• kotlin
• mongodb
• react
• typescript