Can AI Startups Compete with Tech Giants?
If you happen to be a world-famous futurist, inventor, and entrepreneur, what would compel you to take a corporate job? That was one of the questions that I discussed with Ray Kurzweil recently at Synergy Global Forum in New York. Kurzweil was among a large group of luminaries who shared their ideas on innovation and the future with an audience of 5,000 at the conference’s North American debut at Madison Square Garden. I had the pleasure of moderating a Q&A session with the audience following Kurzweil’s presentation. We’ve talked about a great many things, from his views on strong AI to extreme life extension, but this particular point on his résumé—accepting a position at Google in 2012—piqued my curiosity.
It seemed that Kurzweil wasn’t keen on the offer at first, having come to the Google founders with a proposal for funding one of his latest inventions. But after they explained all of the resources and assets they had around data sets and computing capabilities, Kurzweil found the argument compelling enough to take the first job he’s ever had at a company that he didn’t start. So, why does it matter that Kurzweil—a recipient of the National Medal of Technology and Innovation (the highest honor in technology given in the U.S.) and the Lemelson-MIT Prize (the largest cash prize for invention in the U.S.)—took a job at Google? Kurzweil’s decision offers a glimpse into the sources of innovation in a field that is changing many aspects of our lives already and is poised to bring even more significant changes in the near future.
Go big or go home?
If we take a look at how innovation happens in other hot industries today, we see vibrant ecosystems of players large and small inventing and testing things out in the marketplace and collaborating in all sorts of interesting ways. Startups thrive independently and in symbiosis with large companies, sometimes literally next door. Close to home in Kendall Square, the life-sciences capital of the world, established pharmaceutical companies are seeking new ways to work effectively with startups to externalize their R&D for faster and more efficient development of life-saving medicines and therapies. (Admittedly, these paths to innovation pose their own challenges.)
Is innovation in AI different? We keep hearing that for AI to be effective, machines need vast amounts of data on which they can train to become better at whatever task they are intended to perform. This parameter is hard to get around if your company doesn’t have control over massive datasets (like Google or Facebook, for example) or doesn’t have the resources to purchase datasets from companies that amass those for sale. It would seem that the companies with the most data assets and resources at their disposal would have a significant advantage over the smaller market entrants.
At the Synergy Global Forum conference, a lively discussion ensued about the viability of startups in the AI space. Is it possible to compete with the likes of Google, or do you have to be a Google or a Microsoft or an IBM, with the kind of resources they have available, to do this? Kurzweil’s answer was, to some extent, “yes.”
Tech giants bank on massive scale
In 2017, Google declared itself an “AI-first” company. “We are now witnessing a new shift in computing: the move from a mobile-first to an AI-first world,” CEO Sundar Pichai wrote on the company blog. “And as before, it is forcing us to reimagine our products for a world that allows a more natural, seamless way of interacting with technology.”
While Google’s latest commercial offerings may seem modest to some, its AI-focused R&D is impressive. If you were to search the United States Patents and Trademarks Office (USPTO) database of patents granted in Class 706 (Data Processing – Artificial Intelligence), Google is among the top three “patent assignees.” The other two are IBM and Microsoft, unsurprisingly. A recent study by McKinsey Global Institute reports that “Tech giants including Baidu and Google spent between $20B to $30B on AI in 2016, with 90% of this spent on R&D and deployment, and 10% on AI acquisitions.” A staggering number indeed, compared to the $6B to $9B investment by startups, according to the study.
Bigger may not necessarily be better
During our conversation, Kurzweil acknowledged that there are ways for much smaller companies to collaborate and get some of the benefits of scale that the larger companies enjoy. (He cited the Open Source Initiative as an example.)
Even on their own, startups can be quite successful. My MIT Sloan colleague Thomas Malone, professor of information technology and director of the MIT Center for Collective Intelligence, points out that there are plenty of opportunities for innovation and discovery in the field of AI that don’t require access to gigantic amounts of data. He draws an analogy with another industry that requires massive resources and scale: “If you are in the steel industry and you don’t have a large factory, it doesn’t mean that there is nothing you can do. It means that you should do other things,” he says. “You can work on shaping steel that is produced elsewhere. You can work on making the factory more efficient, so that factory owners would buy that service from you. I don’t think it is hopeless at all for startups.”
Last year, Fast Company published a list of the 10 Most Innovative Companies In AI/Machine Learning. And while it’s not surprising that the top three spots are occupied by tech giants—Google, IBM, and Baidu—the list also features Iris AI, an Oslo, Norway-based company with a staff of eight, Descartes Labs in Los Alamos, New Mexico (staff of 25), and an Israeli company, Zebra Medical Vision (staff of 21). These and other small companies made Fast Company’s list because of their unique contributions to AI-enabled technological innovations in a wide variety of fields, from healthcare to agriculture to legal contracts.
We still need crazy ideas and lots of them
“One thing that small companies can often do better than big ones is try a whole lot of crazy ideas and see which ones work. The market lets different small companies try different ideas,” says Malone. He explains, “It’s often harder for a big company to let a whole bunch of different people in the company try to do the same thing in competition with each other. It’s unlikely that they’d all get a fair chance.”
The big companies realize that, of course, and like many of their peers in other fields, they are trying to figure out how to work with startups more effectively. A few weeks ago, Microsoft Ventures, the venture capital arm of Microsoft, announced a global startup competition called Innovate.AI. Google’s Launchpad Studio is a startup incubator focused on accelerating applied machine learning in healthcare and biotech, to start.
May the best minds win
As companies big and small race to capitalize on the AI potential, the billion-dollar question is: who will power all these efforts? Until machines are smart enough to run themselves (and possibly, us), all these companies need people and many of these people will come from academia. The lure of lucrative industry jobs may prove a strong incentive for AI experts to leave the lab. Already there is a growing concern among some academic institutions that industry is going to vacuum up AI engineers and scientists, so no one will be left to teach the next generation of students, let alone advance basic science.
The apparent AI-skills shortage puts engineers and computer science researchers and graduates in these fields in an enviable position to pick and choose where they work. However, it’s important to remember that many bright minds choose academia over industry, driven by the intrinsic motivation for scientific discovery. After all, the hottest applied research—such as that done by the R&D departments at corporations—relies on basic research that may have been conducted in university labs decades ago. (Case in point: MIT’s own Marvin Minsky built the first “learning machine” in 1951.)
Although the “brain drain” issue is real and shouldn’t be dismissed outright, being at MIT gives me confidence in academia as the wellspring of innovation. At MIT’s Computer Science and Artificial Intelligence Lab (CSAIL), leading AI experts are working to apply their findings to solve real-world problems. Just last year, for example, CSAIL researchers teamed up with Massachusetts General Hospital and Harvard Medical School to develop a model for improving breast-cancer detection with the help of machine learning, and designed a virtual reality (VR) system to help manufacturing workers operate machinery remotely.
All this goes to say that, while tech giants do have vast resources and capabilities, important innovations in AI are just as likely to come from a startup or a university lab. At least until human brains are no longer needed, which is a subject I aim to explore more in the near future.