Blame it on Iron Man, that genius, billionaire superhero with a most effective (and virtual) sidekick, Jarvis. The fictional, artificially intelligent assistant has helped spur the imaginations of Silicon Valley’s billionaires in recent years, many investing dearly in AI services as the market warms up to Amazon Alexa, Google Allo and IBM Watson.
Yahoo Inc. co-founder Jerry Yang is perhaps one of the most active in this elite investor club, as his venture capital firm AME Cloud Ventures counts more than a few AI projects in its portfolio. Among them is Ozlo Inc., a self-described integrated knowledge platform with a high-profile team and a strong desire to influence every intelligent machine on the planet.
Founded by Charles Jolley and Michael Hanson, the former head of Android platform development at Facebook and principal engineer at Mozilla Corp. respectively, Ozlo is three years in the making. Today the startup unveiled its open index of knowledge, now available to the business world for training the likes of chatbots, digital assistants and intelligent machines. With $14 million in capital raised from Greylock Partners and Battery Ventures in addition to AME, Ozlo’s extended development time has built up serious expectations for its futuristic vision of AI.
Released as a set of three application programming interfaces, Ozlo’s knowledge index contains over two billion data points on topics ranging from movies to restaurants and bars to the smart home, and specializes in real world interactions. The APIs cover data extraction, intent recognition and conversion actions, all designed to understand and close the deal on business-to-consumer transactions. Instead of narrow, fact-based responses to user queries, Ozlo’s spectrum view of relational data hopes to provide more practical answers to everyday questions.
To give an example, a request to Apple Inc.’s Siri assistant for nearby restaurants that are good for date night merely returns a list of local venues. Ozlo, on the other hand, will present context around nearby restaurants to aid the decision-making process. For instance, Ramona’s is cozy, but gets crowded and loud on weekend evenings.
“This is really the big unveiling of our grand plan,” said Jolley. “We started three years ago with the idea the world would need a new data layer, and you can’t just come up with one overnight. Intelligent systems need this to work robustly. There aren’t places a business can do that. If you want an index that’s built from intelligence from the ground up, we are unique in that sense. Today we can start to deliver on the promise of these intelligent systems… what things mean at the level humans are speaking.”
Ozlo is also more persistent than its rivals, presenting a dynamic array of actions at each step instead of the fruitless “I didn’t get that,” or “here’s what I found on the web” responses typical of less cognitive digital assistants.
Similar to other offerings in the market, Ozlo does look to the web and a plethora of other APIs from sources such as Yelp, IMDb and nutrition watchdogs for its information. Deviating from the standard keyword or popularity-based ranking systems, Ozlo applies its own weighted rankings to web content, sussing out what it calls assertions. Normalizing data in this way, Ozlo can determine that a web page with four paragraphs, each making a unique assertion about a different movie but all referencing the same director, is a professionally written article reviewing four films by a single director. These broken-down data points are then sent to another machine learning system within Ozlo to study the relational interactions with other data points, building up a confidence scale for the answers Ozlo provides.
“One thing we found is that by combining authoritative content with crowdsourced information and professional data, we get different views on the same entity,” said Hanson. “If you look at what a business says about itself, you get very different data than reading reviews on the internet, and different data still if you read professional copy about a business. Depending on the attribute, you’d like your system to be smart enough to pick and choose where the responses give the best interaction. We have the ability to tune those weighted scores specific to the user’s application.”
It’s this confidence-imbued data that’s exposed to Ozlo’s freshly released APIs, drawing back the curtain on Ozlo’s business model in the process. The Ozlo chatbot app, which has been available to consumers for the past year, has been interfacing with humans seeking local restaurant and movie information. In setting up lunch deliveries, movie and date nights, Ozlo has learned from each interaction. With new features also launching on its consumer-facing app today, Ozlo moves deeper into the smart home with interactions for entertainment and cooking.
It’s bigger than the app
While Ozlo’s initial use case seems small, the startup looks to do much bigger things with the relational web of context it’s built. The very act of observing human intent specific to geospatial environments can help train AI how to understand humans. That makes Ozlo’s business offering less about the data points themselves and more about the context a human needs for that data at any given place and time. If comedian Steve Harvey’s empowered the single woman by teaching her to think like a man, Ozlo’s empowering the machine by training its neural paths to think like a human.
While the AI market is projected to reach $16.06 billion by 2022, AI as a software service layer is today a relatively new market opportunity facing many questions about its profitability. Alongside the growing popularity of services like Amazon Alexa comes the commoditization of AI software, and the lowering of both costs and barriers to deployment. With Google Inc. and Facebook Inc. being just two of the tech giants open sourcing portions of their own AI efforts, cognitive machines can be assembled for a variety of business purposes faster and more seamlessly than ever before.
For Google, Apple and Amazon.com Inc., it’s easy enough to understand the business motives behind their digital assistants. Tied to their search data, application software and retail outlets, digital assistants become extensions of corporate marketing efforts and customer service relations. Yet these very reasons could make users wary of corporate-powered digital assistants.
“I understand how gathering all this information serves the business, but how does it serve you?” asked Peter Burris, chief research officer and general manager of Wikibon, owned by the same parent company as SiliconANGLE. Digital assistants and AI are going to be important, he said, “but we don’t know how they’re going to make money — by serving users or advertisers? The use case is still very clunky.”
Ozlo hopes its independent status, untied to the major players pushing digital assistants, convinces users of its mission to put their needs first. Without search ads to prioritize or products to sell, Ozlo is free to shift the paradigm on machine-human interactions. But will that change as the AI as a service market gets more competitive? And how effectively can Ozlo interface with the mainstream digital assistants if their corporate overlords limit third party access and control?
As far as Burris is concerned, Ozlo and similar services are all just remote controls on an end user’s mobile device. Until cognitive machines can truly live up to the hype, he wonders if the tradeoff is worth giving up so much personal data in exchange for one less link to click when searching for date-worthy restaurants.
“Ozlo’s could ultimately say to business clients, ‘Hey, use our APIs. We can pass messages to Siri so it takes commands from us, and tell the Amazon Firestick what to do,’” Burris said. “In the meantime, what are you giving up? An enormous amount of information versus pushing an extra button.”
In the end, it’s the user experience that will evolve AIs from query runners to conversationalists, and Ozlo wants to be the biggest disruptor in this space. Despite the excitement over chatbots and digital assistants this past year, the market is not yet ready for prime time and relies heavily on the mainstream ubiquity of Amazon, Google and Apple services.
Ozlo investor and former Mozilla CEO John Lilly detailed the obstacles still challenging the AI as a service market, admitting digital assistants don’t yet live up to the hype. “They need to understand more and reason better [both areas of Ozlo’s focus]. And precisely because they’re ubiquitous, sometimes they blend into the background so that we forget they’re there. Once they get better at understanding, and are all around us, I think we’ll see usage skyrocket.”