Building an innovation ecosystem, together
Philippe Beaudoin Philippe Beaudoin
July 31 8 min

Building an innovation ecosystem, together

The great American inventor Thomas Edison is credited with many advancements, including the light bulb, the phonograph, and the movie camera. But his greatest achievement, and his greatest failure, provide the two most important lessons for innovators today: innovation is the outcome of a great process, and nobody can do it alone.

Edison’s greatest achievement was much simpler than his inventions, and wasn’t powered by electricity: he put the smartest people he could find together in the world’s first industrial research lab. Many of the ideas for which he is remembered originated with the researchers in that lab, where Edison built a process of innovation. That’s the process we can see today in the best companies working to push science forward.

Edison’s greatest failure was in stopping that process at the company’s front door. He aggressively pursued patent litigation, and shut down many rivals working to improve on his innovations. In time, one of Edison’s inventors moved West to escape the ever-present lawsuits, and set up shop in a bucolic neighbourhood of Los Angeles known as Hollywood.

At Element AI, our innovation pipeline reflects the key lessons from Edison: it’s not about a single idea, it’s about a continuous process of generating and testing new ideas. At the same time, we can’t do it alone, as a single company. We’re building an innovation pipeline as part of a complex, emerging AI network, where the best ideas are put forth and challenged by people across industry and academia. The key to innovation is a healthy, thriving ecosystem: where humans are connecting with humans around great ideas.

The innovation pipeline

The innovation pipeline.

Element AI is working at the forefront of the machine learning revolution, and this time is unlike any that has come before. For the researchers who have worked in AI, it’s almost like watching a dam bursting: groundbreaking ideas are coming up at an incredibly fast rhythm and, maybe more importantly, are moving from research to production at an unprecedented speed.

The best way to take advantage of the breakneck pace in machine learning is building a pipeline for innovation, a continuous process through which new ideas come up and are implemented.

There are four stages in the pipeline: identifying, understanding, productionizing and scaling the best innovations in a continuing fashion.

It starts with identifying promising ideas and figuring out which ones to explore further. Then it's about understanding them better. It's not enough for a few researchers to get it: everyone involved in implementing the innovation should understand it to some extent. Then the idea will need to be productionized. If we're talking about software, it means every aspect of software engineering: high quality code, testing, continuous deployment, and so on. Then it’s about scaling up, deploying the software, maintaining it and ensuring the customer remains happy.

This innovation pipeline then feeds into the wider ecosystem. Yet we need to be clear that the product part of the pipeline is just as important as the ideas. It’s about commercial and academic impact, and how they reinforce one another.

Building an ecosystem

Building and ecosystem.

Even though we're talking about the innovation pipeline as a process to help an idea mature, it’s really about the people. It's humans connecting with humans around these great ideas. To do it alone, all within one company, is bound to fail. This journey is too complex for a closed group to be able to do this on their own. You need a rich and healthy ecosystem.

The first two stages of the innovation pipeline, identifying great ideas and understanding how they can be used, are precisely what the academic world has been historically good at. Researchers have honed their intuition to be able to identify a promising idea quickly and to push its exploration in the right direction. Researchers also happen to be excellent communicators, as a lot of their work depends on their ability to communicate their ideas effectively to their community.

Since researchers are so important to the innovation pipeline and the ecosystem, it's useful to zoom into the academic part of the ecosystem. Take Canada, for example.

In 2017, NSERC, the leading scientific funding agency in Canada, gave grants to 11,210 scientists, compared with 17,847 scientists supported by the equivalent American agency, the NSF. Considering the relative size of the two countries, that’s six times more researchers per capita funded in Canada, where more researchers get relatively smaller grants. As a result, we rarely see these huge independent research labs that are typical in other countries. Instead, successful Canadian researchers will tend to connect with their peers over shared interests.

This tendency to create strong connections between researchers is made explicit by CIFAR, the Canadian Institute for Advanced Research, one of the country’s key sources for scientific funding on long-term research. In its original 1982 mission, CIFAR established that it should fund research around teams rather than individuals, that it should foster deep connections by focusing on lasting research collaborations, and that these collaborations should happen across geographic and disciplinary boundaries.

Concretely, each CIFAR program organizes one or two workshops a year where a small number of very well curated researchers get together and discuss ideas that they are all enthusiastic about. Long-lasting collaborations between remote researchers often emerge out of these workshops.

Connecting research and product

Connecting research and product.

Canada has a rich and healthy academic ecosystem. But for anyone looking to move beyond the research lab and build a thriving ecosystem, it’s important to connect academia with the production stages of an innovation pipeline.

Traditionally, the tech transfer offices of Canadian universities have focused on converting ideas into products. A brilliant professor would come up with an idea, the university would help protect the idea and its implementation, and this intellectual property could then be spun out to a new or existing company. Yet in the fast-paced world of machine learning, building an efficient innovation pipeline requires new approaches to technology transfer that foster long-term collaboration. We need to connect humans together.

More and more Canadian academic institutions and businesses are realizing this fact, and are building mechanisms to help create a strong fabric of individual relationships that span across the academic/industrial divide.

At Element AI, we built our strategy around bridging that divide with an effective innovation pipeline. From the early days, we strived to hire collaborative researchers and engineers were respected by the academic world. At the same time, we built a network of brilliant academics who had a strong desire to help a corporation have a positive impact on the world.

The overall results look very much like a rich, healthy ecosystem where people from different horizons and different interests manage to connect in a meaningful way. And it’s not just about Element AI, it’s about our city and our country.

The Montreal ecosystem

The Montreal ecosystem.

In Montreal, we have a thriving AI ecosystem, and it is something that can be clearly felt. Researchers from different institutions interact at local events, and they will even collaborate on publications. In a famous example from our early days, we published a workshop paper with authors affiliated with Google Brain, Facebook, DeepMind and Element AI in addition to academic labs at the University of Montreal, McGill University, and ENS Paris.

The Montreal Institute for Learning Algorithms, one of the three large Canadian AI institutes, is one of the reasons behind this success. Mila brought researchers together and, equally importantly, created a physical hub for AI scientists to visit and work. Now, within one block of Mila, there are researchers from McGill University, University of Montreal, Element AI, Facebook, Samsung and Microsoft.

Yoshua Bengio himself may well have played an important role in building Montreal’s AI scene through his own example, in addition to his role founding and promoting Mila. As is now well-known, Yoshua decided to remain an academic despite repeated offers to join big technology companies. Still, he helped co-found Element AI and has consistently collaborated with industrial partners, helping them see the value of embracing the academic ideals of openness and reproducibility.

Anyone who comes to Montreal and bumps into an AI researcher will hear all about the latest exciting project they’re working on. And it goes farther than that: any researcher will jump at the chance to tell you about all their collaborators, and chances are they will come both from academic and industrial labs. That’s what a rich and healthy ecosystem is all about: humans connecting with humans around great ideas.

Edison and his lab may have invented and improved the lightbulb and the movie camera, but it’s Hollywood that built the motion picture industry. With its innovation ecosystem, Montreal is turning into something like a Hollywood for machine learning.

Philippe Beaudoin gave a version of this speech in June at EmTech Next, an event hosted by MIT Technology Review. Click here for video from the event.