If I said “artificial intelligence,” what’s the first thing that would come to mind? Some kind of robot that could talk and learn on its own? Maybe take over the world?
Let me dispel any preconceived notions then: AI is not just about robots that can have intelligent conversations with you. Over the past several years, developments in AI — and their implications — have extended far beyond the talkative, inquisitive robots you’ve seen in sci-fi movies, and Boston’s tech industry is an excellent showcase for that.
“Boston is specifically well positioned to lead the revolution.”
Just take a look at Boston’s top five AI startup deals from 2016 — according to CB Insights — and it will give you a good idea of the diversity of problems AI is starting to help solve.
Indigo, which has raised $156 million in funding over the last three years, applies machine learning algorithms to a large database of plant genomes to help farmers make sturdier crops that can survive harsh conditions. NuTonomy, which is developing autonomous vehicle software, beat Uber to the punch last year by becoming the first company to start testing self-driving taxis. Cogito’s software can detect a range of emotions from listening to just your voice, and it’s letting call centers provide better customer service by helping employees keep a better pulse on the mood of a phone call.
And that’s just the tip of the iceberg.
“AI is seeping into our lives without people really realizing it,” Sarah Fay, managing director at Glasswing Ventures, said. Google’s search engine has been using a deep learning system called RankBrain since 2015 to fuel better search results. Facebook now uses AI for better language translation. Amazon CEO Jeff Bezos recently said AI will be used to “improve every business.”
While tech giants like Google, Facebook and Amazon have been quietly — and sometimes not so quietly — bringing AI into our lives, venture capitalists have been increasingly putting more money into new companies that are bringing AI into various aspects of work and life. According to CB Insights, VC firms in the United States invested $5 billion in 658 AI-related companies in 2016, a 61 percent increase over the year. That growing interest is reflected in Boston, where new VC firms like Glasswing has popped up to focus specifically on emerging AI technologies.
Vivjan Myrto, founder and managing partner at Hyperplane, another AI-focused VC firm, said Boston is well-equipped for the new wave of AI because of its sprawling number of universities and research institutions. One prominent Boston AI startup that is based on years of research at MIT is Jibo, which is making a home robot that acts more like a companion than an assistant.
There are also older companies like Roomba maker iRobot and speech recognition company Nuance, whose employees and, in some cases, founders go on to start new ventures. One good example of that: Semantic Machines, a startup founded in Newton that is building conversational AI systems to power the websites and apps of large companies. The startup, which is leaning heavily into natural language processing and deep learning, is led by Dan Roth, who founded a speech recognition and sold it to Nuance for $300 million in 2007.
More recently, Myrto said, universities like MIT and Harvard have put an even greater emphasis into incubating and commercializing technologies, pointing to programs like Harvard Innovation Lab and The Engine, MIT’s new venture fund and startup incubator. At the same time, Fay said, AI is becoming a more popular subject in computer science programs. As the Boston Globe reported, MIT’s Introduction to Machine Learning has become one of the most popular classes on campus, with 700 students signing up this year — only 566 seats were made available after an overflow room was created.
“We’re seeing a big change in the mentality of academic institutions.”
“We’re seeing a big change in the mentality of academic institutions,” Myrto said. “For the first time, [they’re] a lot more commercially focused.”
For some Boston entrepreneurs, a background in AI-related technologies and techniques — like natural language processing, machine learning and computer vision — isn’t necessary for venturing into the world of AI. Sometimes it’s just a matter of recognizing an opportunity, as well as having the right combination of technical expertise and a willingness to learn. That’s the situation Rob May found himself in when he sold his data recovery startup Backupify to Datto in 2014.
May said he started to look AI, among other emerging technologies, when he realized he didn’t want to stay at Datto for much longer. May began reading research papers on machine learning. At the same time, he recognized the rise of work collaboration platforms like Slack and HipChat, which weren’t just providing better ways for workers to communicate but also creating giant repositories of data. From those observations, he arrived at a question that would lead to his next startup: “Can you create digital workers from those chat-based data sets?”
Talla, May’s new startup, is part of the wave of chatbots that have become pervasive in enterprise collaboration software, as well as social media platforms like Facebook and Kik.
With advances in machine learning and natural language processing, chatbots like the one Talla is developing are no longer based on an archaic set of rules that provide canned responses to questions asked in a very specific way. Instead, they are able to learn from conversations and recognize when the same question is being asked in a different way.
For example, May said, machine learning and natural language processing have allowed bots like his to recognize that these two questions are basically looking for the same answer: “how do I recover a lost file?” and “how do I find a lost Word document?”
“That’s where rule-based approaches fail,” May said.
As to what Talla actually does: the startup provides chatbots that can automate human resources and IT tasks, reducing the need for human workers to constantly answer the same questions over and over again from different employees. That, May said, can increase the productivity of employees and, as a result, help companies save money.
Productivity is just one category AI can have an impact. Another area is marketing and advertising. One example of that is Ditto Labs, which uses computer vision to help brands identify their logos and products in photos posted to social media, as well as things like the sentiment of any people who appear in those photos and the setting.
But perhaps a more crucial area where AI can have an important impact is healthcare. One of the latest startups to move into that space is Cambridge-based OM1, which is using machine learning algorithms on the vast amount of healthcare data collected by health systems to create better standards for treatments and procedures. OM1’s software, for example, can help physicians understand how a patient with a particular condition may recover after surgery or react to a certain kind of treatment. It can also take other things into account like age, gender, work habits and social status.
By using AI to find patterns in what works and doesn’t work for patients, OM1 could potentially reduce the risk of readmission and the need for costly procedures. “What we’re really trying to do is personalize healthcare and the way to do that is data-driven,” OM1 CEO Rich Gliklich said earlier this year.
But the important companies in the AI space won’t just be ones that process data, Myrto said.
Equally important are the companies that create or provide new sources of data. He points to one of his portfolio companies, Vesper, as an example. While the Boston-based startup isn’t working in AI, per se, the low-power, water-submersible microphones it’s developing could be used as sensors in industrial settings, for example, to collect massive amounts of data that machine learning algorithms can then be applied to.
Another local startup that is putting thought into the gathering of data for AI purposes is Forge.AI. The startup, which is still in stealth mode, is building software that helps businesses process unstructured data — like things said on television or written in a news story — so that it can be used for machine learning and analytical purposes, co-founder Jim Crowley said.
While Boston has a very active startup community and a solid foundation for building AI companies, May said there are a few things that are holding the Boston region back. For one, he said, there are more aggressive venture funds willing to take more risks in the West Coast. That sense of risk-taking on the West Coast also applies to people willing to adopt new technologies faster, he added, pointing to the fact that most of Talla’s early customers are in the San Francisco Bay area.
But Myrto said he thinks the new wave of VC firms like his, along with the flow of talent from universities and older companies, can turn Boston into “an incredible ecosystem again.”
“Boston is specifically well positioned to lead the revolution,” he said.
This is the first story in our AI in Boston series, which is running several stories on artificial intelligence the week of May 15.