Introduction
My journey into the realm of artificial intelligence commenced within Tencent AI Lab. There, I crafted interfaces tailored for algorithm scientists within the algorithm center. Subsequently, I ventured into designing an intelligent question-and-answer interface for patient triage at Tencent Healthcare. This trajectory led me to my current role in AIGC, where I immerse myself in vertical field applications of AI. Witnessing its evolution from a mere concept to a dependable assistant has been remarkable. Along this path, I've distilled several design principles that remain highly practical even today.
Crafting Conversational User Interfaces (CUI)
In shaping CUI, I aimed to imbue AI's responses with a sense of brevity and spontaneity akin to human thought and typing. This concept initially took form during my tenure at Tencent Medical, where technical constraints prevented real-time responses. Moreover, mirroring typical chat interactions, where thoughts are pondered, words are typed, and revisions are made, fosters a sense of trust.
In the design of Lark AI, I upgraded this design to a visualized symbol.
In UTA AI, I upgraded the AI thinking attributes to components with more similar design elements
Transitioning from Closed-loop to Open-ended Questions
Crafting an exceptional guidance experience significantly lowers the user's barrier to entry.
The original intention of the Magic Panel design is to strengthen guidance and understanding in the CUI process
AI Toolbox
Harnessing the strengths of diverse AI tools spurs the creation of novel content, blending inspiration with efficiency.
In the design of the Lark official website, we integrated local user characteristics and utilized various tools to generate a Persona library with diverse local features.
By combining AI-generated content with graphics or videos, we can automatically create the appearance of most natural landscapes, enabling 24-hour real-time generation.
Emotional IP Integration
Mapping IP images to AI models helps users understand the capabilities of different models. Moreover, it imbues the models with the ability to express subtle emotions during the generation process, such as waiting, contemplating, and focusing, thereby encouraging users to engage more actively in interaction.
