Social Media

Light
Dark

Shortwave’s AI-powered assistant lets you ask questions about your email history

Since the surge in popularity of large language models (LLMs), numerous developers have created tools using OpenAI’s GPT or Anthropic’s Claude to assist individuals in composing emails with different tones and styles. Shortwave, an email client developed by former Google employees, is introducing an AI-powered assistant designed to field questions about your email inbox.

One of the assistant’s primary features is its ability, once granted access to your email history, to locate specific emails in response to natural language queries. For instance, you can ask questions like “What was the link to the video the client sent regarding Project Cricket’s issues?” or “Which emails discussed SaaS startups in the past week?”

Shortwave’s AI assistant is now available for all users to test, accessible by clicking on the AI icon in the top right corner.

Due to LLMs’ proficiency in formatting, you can also make inquiries such as “Summarize emails related to Project Cricket from last week using bullet points.” It’s worth noting that Shortwave introduced an email summarization feature back in March.

Additionally, the assistant offers standard AI-powered email capabilities, such as generating drafts in various formats, enhancing your writing, translating email text, condensing conversations, and making emails more concise.

Since Shortwave’s assistant has access to your calendar, you can even ask it to draft a meeting request with suggested times.

Andrew Lee, CEO of Shortwave, revealed that the company employs a combination of several models for the assistant. They use Instructor-xl and GPT 3.5 for searching, Pinecone’s vector database solution for long-term memory, enabling the assistant to have an extensive contextual understanding, MS Marco for re-ranking emails post-search retrieval, and GPT-4 for the final user-facing output.

Shortwave, boasting over 20,000 active users, emphasizes that these models are not trained on any user data.

Furthermore, the company utilizes GPT-based models, allowing users to search for web-based information, such as the prices and screen sizes of the latest iPhones, without leaving the app.

Leave a Reply

Your email address will not be published. Required fields are marked *