Ready for Astra?
Google DeepMind’s Project Astra is a research project aimed at developing a versatile AI assistant capable of understanding and responding to the world in real time. It's designed to be a more intuitive and responsive AI assistant than previous generations.
Problem Statement: How can we design a user experience for Project Astra that effectively translates its research capabilities into a commercially viable product, while addressing technical limitations, economic constraints, and ethical concerns?
Measurements of success: Google is a tech giant with a massive user base across its various products and services. Here's a breakdown of what positive metrics might look like to validate the product’s success:
User Acquisition: 20 million users sign-up within the first 3 months and a 2% increase in new Google Search users month-over-month
Active User Base: 85% of Google Astra users actively use the service at least once a week, with a low churn rate of 2%.
Subscription: A 5% increase in Google Gemini Advance subscribers, with a renewal rate of 85% within 90 days
Let’s get started.
What’s Gemini?
Gemini is a generative AI model that generalizes and seamlessly understands, operates across, and combines different types of information including text, code, audio, image, and video to help people with writing, planning, learning, and more.
How Does It Work?
When you enter a prompt into Gemini, it replies with a response using the information it already knows or fetches from other sources, like other Google services.
Google's Gemini AI services offer a range of capabilities and products. Here are the main offerings:
The business of
Google DeepMind
Google DeepMind's primary business model revolves around research and development rather than direct revenue generation. In essence, DeepMind's business strategy is to be a pioneer in AI research, driving innovation for Alphabet and potentially creating new business opportunities in the future.
The Approach
Inspired by Google DeepMind, I'm designing the UX/UI for Astra, Google's AI assistant technology. My goal is to create a practical and strategic user experience that can be adopted into Google’s commercial products. My research began with these key questions:
Responsibility
Any AI technology requires a steadfast commitment to three core principles:
Practical - Functional - Safe
I’ve included Google’s Responsibility and Governance for greater details.
Full paper of The Ethics of Advanced AI Assistants is HERE
Solving Ambiguous Problems
Astra's developmental stage limits access to concrete data. To compensate, I'm combining insights from published Google research, general AI usage trends, desktop studies, and personal observations. Early findings reveal a growing reliance on AI across various domains, from professional writing to everyday shopping. This trend extends to tackling both specific and complex tasks.
Imagine a Tokyo tourist, lost in the vibrant chaos of Times Square, their eyes wide with a mix of excitement and confusion. Or a local New Yorker, struggling to bridge the language gap with a potential business partner from Seoul. These are the everyday challenges real-time AI assistance is poised to solve. Envision a world where language barriers dissolve, or information is more readily available, thus fostering deeper connections between cultures and communities.
Use Cases
The real world is complex and dynamic. To create more natural and responsive interactions, the DeepMind Engineering team is enhancing Gemini models. Simultaneously, I'm developing use cases for the Astra AI assistant to effectively support young people in real-world scenarios. By leveraging research and data, I aim to design Astra as a practical tool for everyday life.
Use case #1: Reference Information
A person is chatting with a friend and needs to reference important information but can’t remember it. They ask Astra for help to quickly retrieve and provide the relevant details during the conversation.
Use case #2: Translations AT Local Restaurants
A person is traveling abroad and wants to enjoy a meal at a restaurant. They need to translate the menu and also seek recommendations for the best dishes, based on local favorites.
Learnings & Insights
Learning: A 2024 survey found that younger users are more likely to use AI tools like ChatGPT for research. The 18-24 age group leads, followed by the 25-34 group.
Insight: Target younger demographics with user-friendly AI tools and marketing strategies. Develop features and content that appeal to these age groups and use popular platforms to increase engagement.
Learning: when asked U.S. companies' use of AI chatbots and virtual assistants, as of 2024, approximately 46% of the surveyed companies in the United States claim to use artificial intelligence (AI) tools such as ChatGPT, virtual assistants, and chatbots in their activities.
Insight: Develop AI solutions for customer service and business operations, ensuring they are scalable and customizable to meet the needs of current and potential users.
Learning: The AI market is set to grow by 26% in 2025. If the projections about growth in the AI space come to fruition, further manpower will be required.
Insight: Expand your team and resources to prepare for AI market growth. Upskill current staff, recruit new talent, and consider strategic partnerships.
Insight: Enhance the accuracy and reliability of AI solutions by addressing gaps compared to competitors, focusing on language processing and contextual understanding.
Learning: Data suggests that AI has the potential to boost employee productivity by approximately 40% by 2035.
Insight: Invest in AI technologies to automate repetitive tasks and boost productivity, focusing on areas like data analysis and customer interactions.
Key Insight
Astra demands real-time accuracy and adaptability to thrive. AI's growing dominance across industries, especially among younger users, underscores the need for a practical, user-centric solution.
SiteMap
While Astra's development is ongoing, its core functionalities are taking shape. To ensure a seamless user experience, I've designed a preliminary sitemap. By combining insights from the demo and the established structure of Gemini, I've created a framework that prioritizes accessibility, efficiency, and alignment with Astra's unique value proposition.
User Flow
Astra is designed to be helpful in everyday life. This user flow diagram outlines these two primary use cases: A person is chatting with a friend and needs to reference important information but can’t remember it and another person needing to translate a menu and seek recommendations for the best dishes, based on local favorites.
Consulting With The Professionals
To build a product that meets user needs and technical realities, I consulted with engineering early in the design process. By proactively identifying limitations I aim to avoid costly design iterations that may not be technically viable. The goal is to deliver a product that aligns with both user needs and engineering capabilities.
Take a Look at the Demo: To inform the MVP's UX and UI design, I've analyzed Astra's current look and feel. Combined with competitive research, this foundation will guide the development of a new user experience.
Exploration
Astra's conversational interface is the cornerstone of its value proposition. As an AI designed to seamlessly blend into everyday life, interactions must be consistently delightful and intuitive. To ensure a positive experience, we must prioritize a user interface that adapts seamlessly to various environments, making Astra an indispensable companion.
This image is a snapshot of the Astra demo, which I've used as a starting point for my exploration.
For a more tangible understanding, consider the variety of user environments. From brightly lit offices to dimly lit homes, the background must be designed to accommodate different lighting conditions and device screens, ensuring optimal visibility and readability.
Explorations: Comments
AAstra's dialogue must be clear and unobtrusive, especially when paired with its camera-enabled features. By harmonizing conversation and visual context, I can create an immersive and personalized AI assistant.
Explorations: Home Screens
AAstra's dialogue must be clear and unobtrusive, especially when paired with its camera-enabled features. By harmonizing conversation and visual context, I can create an immersive and personalized AI assistant.
Note: needs to be button that allows for audio/camera usage and text usage. The text needs to be readable and I’ll need feedback on how users want the conversation between themselves and Astra is displayed.
I'll need to design a user interface that supports both text and voice input. Taking into account initial feedback and design critiques, I'll also need to create a specific interface for when the camera is disabled.
In each iteration, moving from left to right, I've experimented with increasingly transparent backgrounds. The second image explored how to integrate the background seamlessly into various environments while maintaining readability. User feedback revealed a preference for a more subtle, Gaussian blur effect, as seen in the third image. Additionally, the 'plus' button was changed to 'send' based on user curiosity and a desire to interact more directly with Astra. As development progressed, I introduced 'hold to talk' and a text option, as demonstrated in the final image, which showcases the background in a different location.
Building a Demo
Demo of Astra
With my MVP UI designed, I'm building a prototype to explore edge cases, low-light performance, and other scenarios. This will help me understand the product better and identify opportunities for improvement before wider release.
Let's embark on this journey together! Subscribe to my YouTube channel and follow my Medium blog for updates!