Collision 2019: Three Inspirations and Two Insights

This time last week, immersion into the Collision conference's innovation ecosystem was underway in Toronto. It was an honor to be awarded the opportunity to join the Alexandria Economic Development Partnership team at the event and to be an exhibitor.

One week later, I am left with three inspirations and two insights.

Three Inspirations

Inspiration 1. The AI Revolution Has Arrived: Multiple companies on the exhibit floor were showcasing off-the-shelf AI for a broad range of applications. Pervasive availability of AI utilities can only accelerate the data revolution.

Companies serious about using AI responsibly with their own data or with third party data will need to understand well the different embedded processing functions from these off-the-shelf utilities as well as potential biases within data and within processing models. Different computational routes could generate different answers. Model choice therefore will matter at least as much as data inputs.

At next year’s Collision conference, it would be great to see a demonstration or test of how different AI engines handle the exact same data inputs. Will they all deliver the same data analytics? If not, how do they differ?

Inspiration 2. Collision Invests in Success: The conference invests in the success of the start-up. Innovative programming sought to nurture success for start-ups and entrepreneurs. Start-ups were divided into three different groups based on their stage of development (Alpha, Beta, and Growth). Investors and potential channel partners roamed through the rough-hewn plywood stands seeking the next great opportunity. Small start-ups like ours were provided with impromptu opportunities to pitch their business limited only by their stamina and willingness to stand on their feet for 8-10 hours.

But there was more. A steady stream of programming in the form of small-format talks, bilateral mentor hours, and semi-private roundtables provided start-ups with the skills they need to manage a small and hopefully fast-growing business. Topics ranged from selecting the right kind of team member for each stage of development to picking a name for the company, to managing growth problems.

And yet there was more. The conference app provided a matching service, linking investors to companies for scheduled 15 minute meetings at the venue.

Inspiration 3. Women in Tech Were A Force (Sort of): The official statistics indicate that 45% of participants were female. For an industry receiving such heat for being hostile to women, this is impressive. Every day, I met inspiring, innovative women at the cutting edge of technology. The Women in Tech meeting area was consistently packed. Women were not just in the audience. They could be seen in leadership positions on stage as well. Collision was clearly making an effort to break the stereotypes and be inclusive. It was fabulous.

However, the vast majority of women on stage were leaders in the softer side of technology (marketing, app design, consumer interface, publishing, human resources, healthcare/pharmacology, Edtech, environmental and social causes) and, interestingly, in investor/VC leadership.

Even more interesting was that the 30 or so women on stage in core coding roles were employed by large companies (Microsoft, Siemans, Cisco, Mozilla, Hulu, Reddit) rather than by start-ups. Few of these women were in data analysis/data science. Are established companies more friendly to female leadership than start-ups? If so, the pipeline issues are more worrisome than media stories suggest.

Two Insights

Insight 1. Public Policy Operates as a Parallel Processing Function in Tech: Conference organizers sprinkled the venue with placards and low tech interactive walls asking provocative public policy questions. Brilliant! The answers alone are worth individual blogposts (stay tuned!):

But at cocktail parties and conference floor booths, I would mention the displays and people would ask me where I had seen them. I had to show them pictures on my phone. The orange stickers suggest at least some people were interacting with these questions at the intersection of innovation and public policy, but I never met any of them in real life.

Much has been written about how social media amplifies the worst elements of human nature. But its not just social media. Increasingly, people seem to shy away from in-person discussion of difficult issues for fear that disagreement will jeopardize a relationship. Sometimes, it feels like the only people ready and willing to offer views on difficult issues are doing so from a YouTube-worthy speaking platforms (think: TedTalks, conference stages).

To the extent that people engage, they do so by sharing videos or GIFs, or by leaving anonymous orange dots on walls. This is not dialogue. It is a broadcast in search of validation through likes and shares that creates an echo chamber. Even the most engaging TedTalk is still a one-way lecture. There is no opportunity to disagree or find nuance (much less a middle ground) at the personal level.

My perspective is probably skewed. Maybe these discussions occurred in other areas.

Participating in Collision as a start-up means being immersed in the pervasive effort to find investor support and customer engagement. The opportunity to engage in thoughtful policy discussion is a luxury few start-ups can afford, particularly if public policy is not their core business.

The Data Revolution is Skewed and Binary: Running between investor meetings, small-format start-up sessions, potential customer meetings, and running a booth for one day left little time for participation in the vast number of really interesting speeches occurring simultaneously across nearly 20 different stages (large and small). The speeches I did manage to attend provided insight into how binary the data revolution is.

  • At one pole, we have data on human behavior collected when an individual interacts with a screen. Examples include: shopping, gaming, video/television consumption, media consumption (news, blogs, podcasts, books, magazines).

  • At the other pole, we have data on human behavior collected automatically by devices such as cars, phones, fitbits, and the growing universe of smart home appliances (thermostats, vacuum cleaners, refrigerators, etc.).

The problem is that neither type of data is really about interaction with another human being in the real world. These data reflect individual choices in ways that may or may not be self-conscious. In addition, the data emanates from a unique subset of the population that is sufficiently wealthy to own and use these smart devices. So the question is whether the habits of this subset of the population are sufficiently representative to justify extrapolating conclusions about the population as a whole. I am not sure.

As far as I can tell, the only data regarding the reaction function from human interaction is generated from social media. This should worry everyone in technology and beyond given the way social media seems to serve as an outlet for the worse aspects of human behavior.

If the idea is to train AI on existing data and the only data we have on the human reaction function comes from (i) human interactions with a screen, (ii) passive data collection from smart devices that monitor movement, and (iii) social media interactions, the risk is very high that AI training will be deeply deficient. If the idea is to augment the training with human commands, then we need to worry quite a bit about the bias of the programmer.

The data revolution needs to understand and address these issues now, when the data is first collected. I don't know the answers, but I do know we have to start asking these questions now.


It was an exhilarating, exhausting, and inspirational three days in Toronto. Positive energy permeated the conference. People seemed authentically interested in sharing ideas about how to push boundaries and reinvent established ways of doing things through technology . I cannot wait to return next year!