Google launches new program to foster AI startups

Traditional methodologies for launching a startup don't apply to AI-focused ventures, Google says. With Launchpad Studio, now backing four healthcare-focused AI startups, Google aims to build new methodologies.

Transforming tech product management Marty Cagan, Founder and Partner at Silicon Valley Product Group, discusses tech startups, the challenges of tech product management and using AI to make products smarter. Read more: http://zd.net/2z2xN55

Now that Google is an "AI-first" company, it's applying that focus to Launchpad, its accelerator engine.

The Silicon Valley giant on Wednesday announced the first four startups that are joining Launchpad Studio, a new six-month program designed to help companies work on specific projects or proposals that incorporate AI and machine learning into their businesses. The first companies selected for the program all come from the health care/biotechnology sector.

"One of the main hopes here is that we can bridge the gaps between traditional industry and the technology industry a bit more," Malika Cantor, program manager for Launchpad Studio, told ZDNet."When we talk to the health care [industry], there's a lot of discussion about big data and machine learning and AI, but it all feels a bit 'hypey' at this point. What we're really hoping to do with this program... is to make all of this real, to really make an impact in industry with some of these cutting edge techniques."

The first four companies joining Launchpad Studio are:

  • Augmedix, a company applying deep learning and natural language understanding to its platform that helps doctors maintain electronic health records (EHR).
  • The company BrainQ uses advanced machine learning and signal processing tools to develop personalized treatment protocols that can help paralyzed people to move their limbs again.
  • Byteflies plans to use machine learning to manage and exploit data streams from its medical wearables, making them more useful for clinical trials and value-based health care delivery.
  • CytoVale is using machine learning and computer vision for the early detection of sepsis, which research suggests kills more Americans than breast cancer, prostate cancer, and AIDS combined.

By working with startups focused on leveraging machine learning, Google can showcase to the broader enterprise the potential range of AI applications. It can also gain insight into how its own products could better serve companies interested in leveraging AI.

"We're trying to figure out what are the tools and platforms we need to build on our end to best support these specific ecosystems," Cantor said. In the health care sector specifically, she said Google can learn, "what are specialized APIs, or certain levels of certification or other tools -- even outside of [Google Cloud Platform] -- that we need to build to support this ecosystem."

Along with gaining new insights about its own tools, Google is aiming to help the startup world develop new methodologies that more applicable to AI-focused ventures.

"We have a hypothesis that the Lean Startup methodology -- failing fast, pivoting, a lot of methodologies that have worked very well for startups for the last couple of decades or so -- don't really work for companies leveraging machine learning," Cantor said.

Machine learning, she explained, is "extremely capital intensive and time consuming." To build a machine learning model, a company has to acquire data, label it, clean it, then train a model and test it out -- all sunk costs.

More traditional tech companies, by contrast, "can effectively sell a product on the front end, gauge interest and build something on the back end," Cantor said.

For both startups and Fortund 500 companies, she continued, "there are huge capital costs and time involved with actually successfully applying machine learning to a product. We're hoping to shorten this process."

For now, the Launchpad Studio is effectively "brute forcing the process," Cantor said, by supplying companies with resources and technical expertise. "But the hope is to extract product methodologies that can be useful for the startups in the broader ecosystem globally."

Augmedix, founded about five years ago, provides one model for machine learning startups: It launched as a more traditional technology services company that provided a telemedicine platform for doctors offering remote care. Now that the company supports millions of patient visits a year, it has data it can use to build a machine learning model that will make the management of EHRs more efficient.

Still, Augmedix co-founder and COO Pelu Tran said transitioning to an AI-enabled company would be difficult without Google's support.

"Our expertise is in building out enterprise salesforces, building strong relationships with health care systems and ensuring a strong degree of reliability for what is a mission-critical service for doctors," he told ZDNet. "They're not the same skill sets you would have if you were to build a machine learning/AI company from the ground up... We're looking for a rapid education in this space, to gain access to resources [from Google] and to the broader marketplace and ecosystem of tools."

The companies that join Launchpad Studio receive equity-free support, access to Google mentors, Google Cloud Platform and G Suite training, credits, and potential access to hardware and datasets.

While it's starting with health care and biotech companies, Google plans to announce the next track for Launchpad Studio in the next few months. The program will continue to accept applications in the health care space on a rolling basis from here on out.

Previous and related coverage

    Google picks four Brazilian startups for accelerator program

    The ventures will take part in a six-month mentoring scheme by the web giant, which includes a two-week stay in its Mountain View campus.

    Google bets on AI-first as computer vision, voice recognition, machine learning improve

    At Google I/O, CEO Sundar Pichai said that all of the company and its products are being revamped to be AI-first. The shift may be bigger than mobile computing.