TLG creates immersive experience with AI to connect nature, technology and reflection

TLG – The Long Game, a future-oriented company, was looking for a way to manifest its concept of “BioSingularity”, where technology and nature coexist harmoniously.

TLG's main challenge was to create an environment where users could reflect on the connection between technology and nature in a deep and personalized way. To achieve this, the platform needed a system that could identify the user's profile and, based on this, generate questions that stimulate reflection, making the user think about the impact of technology in relation to the natural world.

MadeinWeb developed an innovative solution using a system of AI agents that interact with the user to provide a unique experience. The interaction process follows some main steps:

• Initial Interaction: In the first three or four interactions, the language model (LLM) collects sufficient information from the user. From these interactions, the system begins to understand the user's profile.

• Personalization of Responses: Once the user's profile is defined, LLM begins to search for relevant information, based on this profile, to formulate questions that lead the user to reflect on the central themes of TLG — the connection between nature and technology.

• Agent System: To improve the experience, a main agent interacts with the user. After collecting enough information from the conversation history, the agent uses a tool to retrieve specific contexts, containing pertinent data, in order to formulate questions that make the user reflect on their own initial inputs. This creates a sense of connection between humans and nature.

Data processing was developed together with TLG, keeping the company's strategy at the center of the process. TLG provided a database of texts on various topics related to the philosophy and history of technology and nature. To optimize the use of this data:

• We divided the data manually into topics, as each topic was central to personalizing the user profile.

• The input data was embedded in the system using the topics as metadata. Therefore, in the data retrieval phase, the system filters the information based on the metadata extracted by the LLM, ensuring that the model only receives the context relevant to the user's profile. This improves performance and reduces costs by minimizing the number of context tokens used in the prompt.

Amazon Bedrock: We use Bedrock to access the Claude3-Haiku text generation model and the Titan Embedding V1 Embedding Model. The choice for Claude3-Haiku is due to the low complexity of the task, which is to answer questions and assume a profile dynamically, according to the profile of the user who is interacting.

• Immersive and Personalized Experience: The system provides users with a unique experience, with personalized interactions based on each person's profile, leading them to reflect on the relationship between technology and nature.

• Cost Optimization: The use of metadata in information retrieval significantly reduced the amount of data processed by the LLM, optimizing costs and improving the performance of the platform.

• Customer Strategy Integration: Data processing in collaboration with TLG ensured that the company's strategy was reflected directly in the user experience, creating an authentic connection between content and brand philosophy.