Bard Tender is an AI-driven activation that leverages the PALM2 Model, accessible through the Maker suite, to craft a unique selection of signature cocktails tailored to user-specific contextual inputs.
Once the cocktail recipe is curated using PALM2, it undergoes parsing and is subsequently channeled to Imagen, a Google-powered diffusion model accessible via Google Cloud Vertex. This process generates an image representing AI’s conceptualization of the drink’s appearance.
Subsequently, the cocktail recipe is transmitted independently to both the user and the bartender. This allows users to conveniently place their drink orders directly through the customized email they receive and subsequently pick up their beverages at the bar.
My Role
- UX / UI
- Front End Dev
- Back End Dev
- Palm2 Training / Integration
Featured At
- Cannes 2013
- Google IPG Dinner
- Google Food For Though 2023
Design Challenge
Problem Statement
The Event organizers are tasked with creating an innovative presentation for Google events. They need a user-centric and engaging experience that highlights the capabilities of the AI because they want to captivate the audience and showcase the model’s power in a fresh and exciting way.”
Goal Statement
Create an interactive engagement featuring the Palm2 Large Language Model that will enable event organizers to engage attendees effectively, ultimately making an impression on the Google event audience. This will positively impact the event’s success by enhancing audience engagement and understanding of the AI capabilities in a new way. Effectiveness will be measured through attendee feedback, participation rates, and post-event interest in the solution for other events
The Solution
We created an AI bartender that takes the following data; Thematic questions relative to the particular event, List of ingredients available, List of Mixers available.
It generates a personalized cocktail recipe along with an AI-generated companion image, crafting a unique drink tailored to the user’s preferences. Additionally, it provides step-by-step instructions on creating the cocktail and weaves a narrative around the ideal experience of savoring it.
Needs
- Needs To appeal to various levels of technical understanding.
- Needs to be fun and whimsical in some way.
- Need to personifying the event where it is featured.
- Needs to be platform independent
- Needs to run on multiple terminals in a small space
Features
- Deployed in the cloud so it can run on multiple instances.
- Mobile first responsive design fits iPad laptops all available screens.
- Drink delivery via email ( to elevate wait times ).
- Bartender email an drink ID for more efficient process.
- Web App front end ( Angular ) front end.
- Single question per page design to alleviate choice paralysis.
Project Outcome
The project has been showcased at more than 5 events, each attracting between 100 and 400 attendees, including numerous C-level executives. It has received highly positive feedback from all participants and is steadily building momentum, generating increasing interest from diverse Google-affiliated groups. Over 400 unique drink recipes have been generated in the rollout. See below for some samples.
