New ML Startup on a Mission to Improve Video Conferencing

Above: Andrew Blumenfeld, co-founder of Telepath, and Lily Adelstein, creative project manager, discuss Read AI.

One of the biggest shifts in daily life brought about by the pandemic has been the widespread adoption of video conferencing technology. Though it was around long before COVID-19, the acceptance of video calling only now has become a standard method to conduct everything from business meetings, to trade conferences, to friendly chats. While some office places are returning to pre-pandemic norms, there are many who are betting that this shift will endure.

One such company is the new startup Read AI, which just left stealth mode with a funding announcement. The premise of Read AI is that Zoom (and other video conferencing) meetings are here to stay, and that technology can be applied to make those meetings more productive and engaging. They highlight their product’s ability to track your meeting in real-time and provide attendees with a dashboard that helps them understand how the meeting is going. Specifically, Read’s dashboard captures such things as each participant’s total talk time, an analysis of the sentiment of each participant, and other indicators of engagement. To inform this dashboard, Read is relying on the meeting’s audio and visual data for each attendee. 

Artificial intelligence looms large in this startup, as the “AI” in Read AI would suggest. Advances in computer vision make possible the kind of visual analysis that is required to begin to understand patterns in facial expressions, eye movements, and other gestures that could indicate engagement (or lack thereof), as well as positive or negative sentiments. The same is true with regards to natural language processing, and the ability to pick up on clues about how attendees are really feeling about the meeting, based upon their word choices, syntax, and overall levels of participation. 

Use cases offered by Read AI include sales calls during which a sales representative may appreciate the extra insight as to how their pitch is landing with a new prospect, so he/she can adjust on the fly. Read AI also says 53% of meetings have 7 or more attendees, which can make it difficult for a participant to see all the other participants on their screen at once. This lack of visibility can hurt our ability to pick up on visual cues from every participant throughout the meeting. Here, too, Read AI can help by keeping an eye on everyone, even when a human can’t. 

But one might also imagine that for the consistent user of Read AI, it’s technology could ultimately prove very useful beyond helping to navigate terrain with new or many participants. That user may soon build up enough data to help train a highly personalized model that is able to provide feedback that is very specific to that particular user. Perhaps using certain words tends to lower the engagement levels of the person you are meeting with. Perhaps certain gestures are increasing the positive sentiment of your meeting partner. Knowing these things could prove extremely instructive.

This type of development is likely typical of a forthcoming wave of AI-driven tech startups. The rise in video conferencing is an extreme example of a digital transformation, due to the pandemic. But it is part of a significant general trend at speeding towards the digitization of once difficult to accumulate data. Just as it was once unlikely that a user would get even infrequent heart rate readings, now anyone with a smart watch has that data tracked near-constantly, giving rise to a whole new world of potential machine learning applications that use heart rate as an input. The same is true in the world of meetings. Previously, only a very small slice of meeting data was captured-- most often through rough note taking, with verbatim transcripts or recordings a rare exception. Now, suddenly, an enormous percentage of everyday meetings (and for some companies, 100% of all meetings) are happening on a digital platform that can capture every glance, every word, every pause as a data point.

As with most discussions about data, this obviously raises privacy questions. But it also contributes to the playground of digitized data that is growing far faster than we are currently able to make use of its full potential. Read AI is the latest company to try, tackling the new daily reality of meetings for so many people. Here’s hoping they can remind people to unmute themselves before they start talking.