ideas made fresh daily

Unified-IO

A new general-purpose model with unprecedented breadth, Unified-IO can perform a wide array of visual and linguistic tasks.
 
Diagram outlining how the AI model Unified IO works, by inputting imagery and text, and outputting generated images and text

Unified-IO

Skills: product design, ux design, ui design, identity and brand design

Situation:
PRIOR (Perceptual Reasoning and Interaction Research) is a team at the Allen Institute for AI that works on computer vision research. In other words, PRIOR researches how robots can “see”. The PRIOR team approached ReViz in early 2023 with an idea for a new interactive demo. The team had a model capable of a broad range of text and vision-based tasks, but weren’t sure how to showcase them both to everyday users and the broader research community and did not have a product vision in mind. Additionally, we had a week’s turnaround to launch.

Action:
I managed the product strategy right away, ironing out what tasks we’d like to show, and how to organize the many examples of this model’s capabilities in a way that was accessible to an everyday user, yet encompassed a more AI fluent audience as well.

This process lasted from an initial outline, to a wireframe, to a high-fidelity mock wire, all the way to final launch. On top of managing the content outline, I was responsible for quickly engineering the UX in a matter of 2 days. The interface was also quick turn, which allowed us to go from initial meeting to launch in a week. Alongside the UX and interface, I also designed the identity suite, and fitted it into a press kit for the demo’s launch.

Links:
Unified-IO

 

Unified-IO identity development