Shortwave, an email client developed by former Google employees, has introduced an AI-powered assistant that enables users to ask questions about their inbox. The assistant utilizes large language models (LLMs) to assist in email organization and retrieval.
Shortwave’s new AI-powered assistant revolutionizes email organization by allowing users to ask natural language queries about their inbox history. With features like advanced search, formatting, summarization, and integration with other AI-powered tasks, the assistant enhances productivity and efficiency for email management.
Find Emails with Natural Language Queries
Shortwave’s AI assistant allows users to find specific emails from their inbox by asking natural language queries. Once granted access to email history, users can inquire about emails based on particular content or topics. For instance, one can ask, “What was the link to the video that the client sent me discussing Project Cricket’s bugs?” or “Which emails discussed SaaS startups in the past week?”
Users can begin exploring the capabilities of the AI assistant by clicking on the dedicated AI icon located on the top right corner of the Shortwave email client.
Advanced Formatting and Summarization
Thanks to the proficiency of LLMs in formatting, users can also make queries that involve email summarization and formatting tasks. For example, one can ask the assistant to “Summarize emails about Project Cricket from last week in bullet points.” Shortwave has already launched an email summarization feature earlier in March.
Besides helping with search and summarization, the AI assistant is equipped with various other features commonly found in AI-powered email tools. These include composing drafts in different formats, improving writing skills, translating text within emails, summarizing conversations, and making emails more concise.
The AI assistant can also access users’ calendars, allowing them to inquire about scheduling a meeting draft with proposed timings.
Underlying Models and Data Privacy
Andrew Lee, CEO of Shortwave, revealed that their assistant utilizes a combination of different models. It leverages InstroctorXL and GPT 3.5 for search functionalities, Pinecone’s vector database solution for long-term memory to provide contextual information, MS Marco for email re-ranking after search retrieval, and GPT-4 for final user-facing outputs.
Shortwave emphasizes that these models are not trained on any user data and ensures privacy protection.
In addition to email-related features, Shortwave integrates GPT-based models to facilitate web-based knowledge searches within the app. Users can inquire about the latest iPhone prices and screen sizes without leaving the Shortwave email client.