Once I showed them what AI could do, there was no going back
TLDR: GC Entrepreneurs are currently wrapping up work on a series of interviews with Artificial Intelligence (AI) practitioners across the Government of Canada (GoC). This exercise to understand the GoC’s AI capacity suggests that though AI practitioners may be spread out and loosely linked across many departments and agencies, they share common victories but also challenges.
—
One of our mandates as GC Entrepreneurs (GCE) is to advance and experiment with disruptive technologies such as AI. Our early discussions at the Task Force on Public Sector Innovation as well as those between GCEs made apparent that though there is a number of well-known groups making use of AI tools such as the National Research Council, Immigration, Refugee and Citizenship Canada and my home department of Employment and Social Development Canada there was no broad understanding of what is the state of AI capacity across the GoC. With this realization, it’s clear that we are going to have a hard time growing the GoC’s capacity in AI if we don’t know what it currently looks like. Being “entrepreneurs”, we identified this as a new market poised for growth and decided to get out there and find out.
We went out with a clear goal. Not one to catalog AI projects, or how much was being spent on AI related activities, but to understand who are the practitioners of AI in the GoC. Where did they get their AI skills? How are they using these skills and to solve what kinds of problems? What challenges are they facing? What supports are they getting? This inquiry would point us to potential paths to better support the development of AI capacity across the GoC.
Now that we had our scope defined (manifested as an interview questionnaire), our next challenge was finding AI practitioners. Through a combination of posting our intent on GCConnex and reaching out through the GCEs considerable networks, we were able to conduct 25 interviews with teams across 20+ departments and agencies representing approximately 350 practitioners (1). This number exceeded our expectations. Though the size of the teams varies greatly (from 1 to 80), there is an undeniable breadth to the take up of AI tools across the GoC.
As a teaser for our upcoming reporting on this initiative, I’ll provide you with some insights into what we learned about one group in particular, the lone AI pioneers. These brave souls are exploring on their own the possibilities for AI and applying AI tools in government (20% of those interviewed). Their first defining trait is their passion for AI. A passion that drives them to learn (self-guided) the statistical methods and computer programing languages needed to understand, code and utilize AI. Another common trait is that they have acquired these skills outside of work or formal education. These skills are learned at home (during leisure) using the vast array of free online resources including courses, academic literature and sharing platforms such as GitHub and freely available software such as R and Python. Another interesting trait is though they come from varying educational backgrounds such as physics, economics or computer sciences, their fields all share a foundation in mathematics and statistics.
Once these pioneers have grown accustomed to AI tools, such as coding a neural network to recognize handwritten numbers (http://yann.lecun.com/exdb/mnist/) they quickly start to see potential applications in their work at the GoC. This leads us to another common trait amongst AI pioneers, perseverance. When asked what was one of the greatest barriers they were able to overcome, they said: time. The time to experiment with AI, the approval to install new software and sometimes get new computer hardware. This might sound simple, the path to the accessing software is long and dull filled with waiting, forms and more waiting. The time to develop AI solutions can also be considerable. However, those I spoke to eventually reached their goal and developed AI solutions for their workplace. Though I’m paraphrasing, the outcome of their efforts was always: “Once I showed them what AI could do, there was no going back.”. The investment paid off.
These lone AI pioneers, who understand their business inside out and can see the potential of AI tools in their work are a critical asset in a department’s first steps towards utilizing AI. They can point us to where we should invest to reap the benefits of AI. But this raises an interesting question, is our current environment to foster the growth of AI a sustainable strategy for identifying opportunities and benefiting from AI in the GoC?
I for one think we have some work to do.
I hope you found this interesting. We’ve learned a lot more so, stay tuned. A number of reporting products will be produced over the summer and will be shared broadly. If you would like to hear more or see opportunities for us to present our findings, do not hesitate to contact us.
Mathieu Audet
GCE
(1 ) — For those interested in the qualitative research aspect of things, though we conducted “only” 25 semi-directed interviews there was a very high degree of convergence (often referred to as interview saturation) in our findings, meaning that there was limited variation in how participants answered a many of the questions even when they differed across other characteristics such as their department/agency, personal characteristics and functional areas they work in. Our findings are also supported by another independent source of inquiry (often referred to as triangulation) which in this case comes from the Survey of Artificial Intelligence in Federal Government Departments and Agencies collected for the Deputy Minister Science and Technology Committee — June 26 2018. These two converging and independently collected lines of evidence provides us with a high degree of confidence that what we heard can be generalized to the broader community of AI practitioners.)
