By many accounts, 2017 is poised to be a breakout year for virtual assistants and the artificial intelligence (AI) technologies underpinning those tools. This year’s Consumer Electronics Show in Las Vegas featured the dominance of Amazon.com Inc.’s (NASDAQ:AMZN) Alexa virtual assistant which was baked into all kinds of consumer product offerings at the show, from refrigerators to automobiles. Wider development of AI and virtual assistant technologies are already being produced by many of America’s tech giants and should support one-fifth of human interactions with smartphones by the year 2019.
As defined by market research firm Gartner, virtual assistants are conversational, computer-based characters which simulate a conversation to deliver either voice- or text-based information to a user via a Web, kiosk or mobile interface. A recent survey of American consumers by sales and marketing AI developer Conversica found that 52 percent of respondents felt that the use of virtual assistants at work would help them to be more productive in their jobs, with younger respondents being much more willing to use virtual assistants than senior citizens. Virtual assistant tech is also making forays into customer relationship management (CRM) operations in industries including healthcare and air travel. Millennials have been embracing the use of virtual assistants for performing a wide range of simple tasks, including making phone calls, playing music, shopping, setting alarms and checking their personal calendars.
Despite the fact that every major tech name is involved in the development of virtual assistants, conversational AI services continue to be introduced by smaller firms and startups in the space all over the world. French telecommunications company Orange (NYSE:ORAN) has recently rolled out a new Djingo virtual assistant service for its customers which is incorporated into the company’s set-top boxes. French information technology services firm Atos (EPA:ATO) has recently released the Atos Virtual Assistant providing machine learning capabilities via cloud services to improve user IT support. In mid-April, Cambridge, MA-based software developer Pegasystems (NASDAQ:PEGA) unveiled the Pega Intelligent Virtual Assistant, an AI tool which uses natural language processing to create responsive chatbots for a variety of business scenarios.
Lucida is an open source platform designed for the development of intelligent personal assistants which incorporate image and language processing technologies. Based on research performed at the University of Michigan’s (UM) Clarity Lab, Lucida began as a research platform for identifying ways in which future data centers and cloud computing technologies should be developed to optimize virtual assistants, according to UM research fellow and Clinc Inc. VP & co-founder Johann Hauswald. “Through open source packages, we’ve actively developed open source tools and put together a representative application to see what such an assistant would look like if deployed by Google or Apple,” Hauswald said.
Hauswald said that the work of Clarity Lab and Clinc, an Ann Arbor, MI-based startup founded by the UM research team, was responsive to desires among business contacts which want to develop their own virtual assistant without relying on Siri or Google Now. “The differentiator here is that we allow full customization and control of not only the features but also the data,” he said. Companies seeking to develop their own virtual assistants want to do so in order to keep their customer data private and not create privacy issues with major tech firms having third-party access to that data.
Clinc is a startup based on the research performed at Clarity Lab but there has been no tech transfer from the university, Hauswald said. Rather, Clinc uses the open source Lucida platform to develop virtual assistants for personal finance applications. Last March, Clinc closed its first round of seed funding, taking in $1.2 million and reaching a market valuation of $5 million. This February, news reports indicated that Clinc raised another $6.3 million in funding which the company would use on additional R&D and hires.
Hound, the virtual assistant developed by Santa Clara, CA-based sound search tech developer SoundHound Inc., is an option on the market which is backed by a fairly strong suite of intellectual property. “We’re the only private company that has the entirety of the technology suite you need to do natural language processing,” said Katie McMahon, general manager at SoundHound. Readers might remember the original SoundHound app first released in 2005 which could find information about songs using sound recognition technology. Hound uses similar technology to perform state-of-the-art natural language processing tasks.
One competitive edge held by SoundHound is its portfolio of what McMahon called “speech-to-meaning” technologies. Most virtual assistants utilize speech-to-text technologies which transcribe what a person says into text and then processes that text for meaning. This introduces an extra transcription step which can introduce errors in transcription and add to the latency of the technology. The deep meaning understanding of SoundHound allows users to query the system using questions which have a much more complicated syntax than what a Siri or Alexa can handle. One example given by McMahon is the voice command, “Show me hotels in Seattle with amenities [within a certain price range] within two miles of the Space Needle.” “When you envision the future in 25 years, are we going to speak stilted Tarzan speak to everything around us?” McMahon asked, noting that SoundHound has gone to excruciating pains to develop technologies which effectively answer the need for speech-to-meaning technologies. “It takes a small, nimble group of folks to go through the painstaking process in R&D,” she said.
Along with the Hound and SoundHound apps, SoundHound also offers a Houndify virtual assistant development platform for businesses to create customized Hound operating systems for their own customers. According to McMahon, the Houndify community included 24,000 developers as of 2017 producing hundreds of active Houndify projects. When Houndify launched in 2015, it already offered developers 100 domains in natural language processing including local search, unit conversion, weather, financial information, site status and places of interest. Comparatively, McMahon noted that Siri launched with six domains and even after years of development, both Siri and Google Now only offer a couple dozen domains.
Although McMahon wouldn’t disclose the size of the company’s patent portfolio, she did say that we can “assume that intellectual property is critical and tantamount to what we do.” The company has been able to effectively leverage its portfolio for successful rounds of investment funding, including a $75 million round of funding which closed this January. That round of funding followed announcements earlier in the month that SoundHound has established Houndify partnerships with audio device developers Onkyo and Sharp for developing their own customized virtual assistants for use in their respective devices. Last September, SoundHound announced that Uber and Yelp were developing their own respective virtual assistants on Houndify for developing voice-enabled conversational intelligence tools.
Sunnyvale, CA-based Mezi offers a virtual assistant which focuses on a person’s travel needs. Mezi actually pivoted its focus late last year from an all-purpose virtual assistant to a travel assistant which, unlike other online travel booking services, handles both discovery and the actual transaction for the consumer. “Through major partnerships with travel category leaders like Expedia and American Express, Mezi is targeting its search and personalization functionality on corporate travel, creating both efficiency for the corporate travel agent and a seamless experience for the end-user,” said Swapnil Shinde, CEO and co-founder of Mezi.
According to Shinde, Mezi is attempting to create the first end-to-end travel booking experience for corporate travel which is entirely powered by artificial intelligence. The virtual assistant utilizes natural language messaging interaction tools which both create a more personable experience and helps to gather requirements and preferences for each corporate employee booking travel options through the service. As of this February, the effectiveness of Mezi’s virtual assistant bot was allowing the company to serve three times as many customers that the company’s human employees serve each day, completing 60 percent of Mezi’s flight requests and 38 percent of hotal requests processed by the company.
Shinde also noted that Mezi can also help business professionals adjust their travel itinerary in the midst of their trips. “Users tell Mezi what needs to be done and then Mezi handles the onerous tasks like re-booking flights, changing hotel reservation, updating dinner reservations and anything else that might come up as well as make personalized recommendations on things to do and sights to see in the area for maximizing after-work hours in a new city,” Shinde said. Last July, Mezi announced a $9 million Series A round of funding months before the company shifted its focus towards travel booking services.