Forbes recently announced the alpha release of Adelaide, a cutting-edge AI-driven news search engine. To delve deeper into Adelaide’s offerings, you can explore the release news page provided.
What is Adelaide’s primary function?
Adelaide facilitates engaging individuals in natural conversations to better understand their preferences and requirements, thereby simplifying the search and exploration process.
The introduction of Adelaide signifies a significant advancement in how media enterprises can enhance market engagement by being among the pioneering major publishers utilizing conceptual AI for news aggregation.
Forbes.com’s Exploration of Generative AI to Enhance User Experience
The initial phase of our exploration into conceptual AI tools involved assessing various potential use cases to enhance the user experience on Forbes.com. While our team was excited about the myriad applications of Gen AI, we swiftly settled on a solution geared towards improving the research experience for users casually browsing the Forbes archive.
Identifying an Optimal Intelligence Tool
Our foremost requirement for the new product was a tool that could blend the speed, reliability, and precision of search interactions with the flexibility and creativity of an expansive language model.
Our team’s primary focus was on finding a solution capable of recommending relevant articles closely aligned with user queries while also being contextually aware.
Ideally, we aimed to deliver a chatbot-like experience to users, presenting articles that allowed them to delve deeper into their inquiries.
It was imperative for us to deeply integrate Forbes’ data into our AI products. Leveraging the collective works of our numerous journalists and contributors was crucial in distinguishing our approach from merely relying on large vocabulary models.
By adopting this strategy, the model could minimize errors and deliver responses that were more socially appropriate across various contexts. This integration of requirements led us to adopt the innovative Vertex AI Search & Conversation tools from Google Cloud Platform.
Seamless Integration of Vertex AI Search & Conversation
The setup process for Vertex Search & Conversation was remarkably straightforward and efficient. Establishing the datastore powering the extensive language model was as simple as defining inclusion criteria, writing, and sharing, given that our team already had data stored in BigQuery and Google Cloud Storage that seamlessly integrated with Vertex.
Furthermore, incorporating content metadata into our datastore significantly enhanced model performance and improved the relevance of presented responses. To kickstart the process, all we had to do was access the Vertex SearchAPI. The user-friendly nature of Vertex Search proved to be a major advantage for our team.
While our focus was on enhancing solution development for user experience and usability, our development resources could be allocated to other projects as needed.
Refinement and Testing of the AI Application
Subsequent to establishing a viable proof of concept, our emphasis shifted towards testing, refining, and structuring the responses generated by our new tool. With the support of GCP and our account management team, we engaged in productive discussions that led to rapid iterations addressing issues and enhancements on both ends. The initial iteration of the tool was made available as an application within our internal communication platform, Slack.
The decision to leverage Slack was intuitive as many non-technical team members were already familiar with the platform, enabling them to explore Adelaide’s functionalities effortlessly. By avoiding the need to rebuild the backend infrastructure repeatedly, we could swiftly iterate on Adelaide’s feature set.
Given our team’s paramount concern for rigorous testing, we were meticulous in ensuring that Adelaide’s responses were appropriate for the vast user base expected to access the tool.
Development of a User-Friendly Adelaide Interface
Our design and backend teams were dedicated to crafting an Adelaide-centric experience characterized by clarity and simplicity, focusing on presenting both the conceptual summaries generated by the extensive language model and relevant articles authored by Forbes contributors.
Additionally, we aimed to provide guidance for users new to this type of research experience, considering the novelty of relational AI technology.
Another key aspect of Adelaide was the ability for users to engage in extensive conversations, posing multiple follow-up queries while retaining the context of their previous interactions.
Unveiling Adelaide and Initial Reception
As the launch date for Adelaide approached, we cautiously anticipated the reception the product would receive. Despite encountering last-minute bug fixes, security updates, and logistical challenges common to new product launches, we successfully introduced the tool to a select group!
The positive response from top leadership and the media was encouraging, with a noticeable uptick in Adelaide usage observed. To ensure an optimal user experience, our team remains vigilant in monitoring the tool and implementing enhancements.
Insights and Future Prospects
This project stood out as a significant endeavor for several team members, garnering considerable attention and effort. The experience gleaned from this project, which mirrored larger internal or backend projects, underscored the importance of applying our processes to projects with broader visibility and impact on audiences.
We gleaned valuable insights into relationalAI, realizing that not every response from Adelaide needed to mimic human cadence or phrasing precisely. Instead of fixating on minutiae like comma placement, our focus shifted towards ensuring a cohesive user experience.
Despite the challenges, working on Adelaide proved to be an enjoyable experience, and we look forward to further exploration of conceptual AI and expansive language models across all Forbes content.
