Some people talk about artificial intelligence as a ‘black box’. At Vubble, there is no such thing. Here’s Mariah Martin Shein, Vubble’s Director of Machine Learning, with an inside view.
Above: Vubble co-founders Tessa Sproule (left) and Katie MacGuire
The Vubble team was driving from Toronto to Waterloo, Ontario recently to attend Communitech’s True North tech conference and while we were on the road, I interviewed Vubble's founders Tessa Sproule and Katie MacGuire on what makes Vubble tick. Here are my notes from our engaging chat:
Why did you start Vubble?
T - We were concerned about the future of public media. Katie and I spent two decades living and working (at the CBC). The idea of media being the filter through which people understand the world around them is very much at the core of how we approach everything. We wanted to find a better way for people to be exposed to information media content that they need to see or is important for our democratic institutions — for people to understand complex issues.
You’ve got to get people engaging with issues and ideas that are outside of their comfort zone, outside their filter bubble, so that they see a broader, richer, fuller picture of the world.
Where did the company name come from?
K - Vubble = video + bubble. “How do we break through filter bubbles that are forming around people's experiences in the digital space” was one the first things we tackled. So video and bubble came together.
How has the company grown since you first launched in 2014 to become a leading Canadian media tech team?
K - We started Vubble as a media company curating videos from across the internet. We had the idea that eventually we could build an algorithm that would distribute those videos in a way that burst filter bubbles. We built a large audience very quickly, but we realized that the media advertising business model would not work for us.
We decided to build a software platform that could be licensed. This is a sustainable model that gives us the freedom we need to solve the problems we are after.
Our first real business breakthrough came when we started working with the CMF (Canada Media Fund). We received funding to build our core AI from their innovation program in 2017.
T - The CMF was a huge advantage to us. If we were an American company, we would probably be in the VC space working from the perspective of how do we quickly build enormous value and sell. Eighty percent of my job would be spent trying to build up the story of what we are, rather than to actually building the tools and solutions that are going to change the world.
Why is Vubble different than the other companies?
T - I think first and foremost, it’s that we have the human element — our editors. A lot of companies are coming at it purely from a technological play. There is great advantage and potential scale in that. But there's also enormous risk, and that’s come to light recently with things like the Cambridge Analytica scandal and the boycott YouTube from advertisers.
No technology is yet able to identify what's actually happening inside of a video with better accuracy than a human. And that's becoming a massive issue with things like fake news, and the deep fakes that are now starting to permeate on the web, where you've got fake video that is believable when you look at it; it’s terrifying, to be honest. The nuance of what a human can pick up on is the thing that separates us from the robots.
K - Vubble is a values-led company and that makes us unique too. Misinformation, disinformation and fake news threaten how we gather evidence-based information on a mass scale. The technology and services we offer are solutions, in part, to this problem. Vubble’s clients are also concerned with this problem and want to be part of the solution.
What’s the biggest business problem you’re trying to solve for media?
T - Content discovery and the erosion of digital revenues. Conventional broadcast and publishers under threat because the old models of discovery and revenue no longer exist. So they're fighting amongst themselves. Everybody is fighting for the scarce, waking hours that people have to consume content, whether it's information or entertainment content. Netflix is right up against us — we're all up against against each other, which is crazy.
Our media clients are attracted to Vubble because we are helping them break through the legacy models that are challenging the ways they do business.
What is your solution for the media?
K - Our media customers own and generate a lot of videos. A lot of that video is not being watched. We help our customers to bring those “lost” videos back to life with our proprietary AI-powered video distribution platform.
We also build distribution tools that help media publishers be where their audiences are: via email, native bots, chatbots, discovery boxes on web pages, or even feeding recommended videos directly into their video players. We are also working with media clients to get a deeper understanding of their audience through new kinds of video data and as a result, to create premium advertising opportunities.
How does it work?
T - We have reverse-engineered how some of the big platforms work with their AI. We asked ourselves: if you can create machine learning that understands how to deliver content based on categories to match ads, can you also do that to get people engaging with content in more meaningful ways as well?
So our machine learning, which is delivered via tools like our Vubble native bot, sends content to consumers in category areas that they are interested in. We also show content that differs from what that consumer would usually receive.
That's what’s really important — you’ve got to show people content that is going to be serendipitous and engage with them in surprising and interesting ways, and they're going to find delight in that experience. This is a powerful way to increase user engagement and keep them inside the publishers’ ecosystem.
What can Vubble do for education?
T - I have kids and my co-founder Katie has kids. They would never watch conventional broadcast and they've never picked up a newspaper. How do we reach the digital citizens of tomorrow? Media companies are getting to them by publishing to YouTube and other digital channels. But as a parent, I feel very uncomfortable sitting my young daughter in front of YouTube, because god knows what she's gonna find in her “up next” recommendations.
Education was a natural byproduct of what we are already creating. We have human editors, journalists, do the hard work of evaluating, vetting, and assessing the best information video worth watching every day, and one obvious place that needs that is the education space.
We curate education feeds based on topics such as STEM, history, social studies, the arts... We provide content that is safe, secure, that provide great and timely value in terms of the information and stories that they tell.
What can Vubble do about fake News?
K - There are great fact-checking services out there, like Snopes, but that's not really the solution. We believe the best solution is to raise critical thinking skills across the board. Our solution is the Vubble Credibility Meter. Our editors use it on every single piece of video that comes through our system. They evaluate the credibility of that information based on a 15-point metric scoring system that we developed at a hackathon at MIT in Feb. 2017. The tool is like a nutrition label for content, we know that sometimes you're going to want to have sugar in your diet and that's okay. You just want to know when you're having sugar in your diet. You need a well-balanced diet of information to be a well informed citizen of a country like Canada, or any democracy.
It’s interesting, because sometimes something from a fairly unknown YouTuber might have a higher score in credibility than something from a conventional media organization. So that's the other thing that's really important too — just assessing it based on the source doesn’t work and that's how some technology platforms like Facebook have been approaching it.
Now anybody with their phone can have the same reach as a network of the past and the playing field is quite unfair. Just because someone is producing video from their basement doesn’t mean that is not as credible. It might actually be a first hand account UGC piece of video that is very journalistically sound.
What’s New at Vubble?
T - We just wrapped the production phase of the CMF and have launched a suite of tools that are built to solve real problems that the Canadian and international media industry are wrestling with. We are developing a native video bot which we think will change the way AI is used to engage audiences.
And we’re excited about a new partnership with Seneca’s Applied Research, Innovation and Entrepreneurship group for an exciting new research project targeted at using AI to advance video identification that will automatically identify the subject area of the curated videos in the Vubble library. We're constantly in market research mode, looking at how people are consuming information digitally. We have our fingers on the pulse of how people’s habits are changing.
We get many of these same questions at conferences and in meetings with new clients. Please let us know if you have some questions of your own. Thanks.