Smart, and getting smarter.
Along with introducing a new way for teams to share their connections, today we are announcing the start of Trove’s public training period.
Over the past few years, we have been quietly building the foundational technology behind Trove to deliver insights about your network from your historic messaging data. Through a combination of domain-specific traditional AI and machine learning approaches plus state-of-the-art deep learning and intelligent assistant interfaces, Trove provides a deeper understanding into the meaning of our communications than ever before.
Trove can show you the origin of your relationships and who has made the most introductions for you. It highlights who is waiting on a response and measures your responsiveness, one relationship at a time. Trove automatically finds and surfaces useful contact data like social profiles, phone numbers, job titles, companies and more. As you communicate, Trove suggests various personality archetypes that people in your network exhibit based on their communication patterns. All of this helps you be more connected and keeps the vast number of people you interact with in context.
These capabilities and many others will all become more accurate and powerful as we train-the-brain behind Trove. Here is some information about how that works.
Why does Trove need public training?
While we are excited about how useful Trove already is, it will be even better with fine tuning. Like all AI, the key ingredients for success are 1) loads of data, 2) tireless research, and 3) ample training. We are proud of how far our team has brought Trove’s AI with over 100 thousand connected mailboxes and our continuous internal and beta user training. Naturally, the next step is to open up the training floodgates directly to everyone who uses Trove. Nothing we do in the lab can compare to the benefits of this type of crowd-sourced training for improving Trove’s accuracy and usefulness.
What is being trained exactly?
Trove uses various deep learning approaches to deliver many of its features, such as determining the response intent of an email, parsing contact details from email signatures, summarizing relevant text, identifying personality archetypes and expertise's of people you know, and more. Another example of deep learning we employ are neural networks used to construct Trove’s responses for our voice and chat assistant across many devices.
Is the content of my email kept private?
Yes. Trove analyzes your email messages, but it does not store them or use the contents in any way other than to provide you a great experience.
During the training period, data you provide will teach the AI to provide better results. For instance, when you tell Trove that it identified a contact incorrectly, the AI learns how to parse information more accurately. Occasionally Trove will misunderstand a question you are trying to ask. Offering corrections will help Trove interpret people better next time.
For detailed information on how Trove protects your privacy, please visit www.trove.com/trust.
How can I participate in Trove’s public training?
The first step is simply to start using Trove, especially on a team that is sharing connections across an extended network. Basic actions within our apps go a long way toward training Trove’s AI.
Over the coming months, we will introduce additional user training tools across all of Trove’s apps. These tools, which will be similar to what we use internally, will allow you to see where contact details originated from and flag any incorrect info. You will also see how Trove’s assistant interprets your queries so you can make corrections in real time.
What do I get out of participating in this?
All of this training makes Trove more powerful for everyone who uses it. Together, we are building the most advanced communications AI there is. Our team sees a bright future for digital communications. We hope you are as excited as we are to create this future, together.