Earlier this autumn DeepMind announced the launch of their second lab outside of the UK in Montreal, headed by Doina Precup, Professor at McGill’s School of Computer Science. The Google owned company is famous for its programme AlphaGo - the first AI to beat a human at the game Go. The Montreal lab will focus on reinforcement learning - Doina’s expertise, and will continue DeepMind’s mission of solving intelligence.

We were lucky enough to be joined by Doina on the Women in AI Podcast when we sat down to chat about her current work and thoughts on the current landscape of AI in Canada. The episode is now available to stream on our website or on iTunes.

Initially, I was interested to hear about Doina’s journey to becoming the pioneer she is today, and she explained:

‘I’ve always been interested in AI because I love sci-fi books and movies, and I saw lots of friendly robots interacting with people doing nice things for them’.

This interest was sparked at a young age and through her education at a computer science high school in Romania. The school had a 50/50 boy girl split and there was no discrepancy - Doina’s grandmother was a maths teacher, her mother was a professor at the university teaching computer science, so she didn’t see a gender gap until moving out of Romania. She explained that when she reached the U.S, ‘the proportion of girls was quite shocking...I’m still trying to work out what drives this.’

Upon speaking about her current work, Doina said  that she’s ‘starting to split time between McGill and DeepMind Montreal. At McGill we have a lab which is co-directed by four professors: myself, Joelle Pineau, Prakash Panangaden, and Jackie Cheung and we do research in reinforcement learning building new algorithms for learning automatic abstractions from data’...Doina goes on to explain the intricacies of this research and how it’s impacting society through the applications of reinforcement learning.

Doina discusses her biggest success working in AI, and shares her work in temporal abstraction - constructing representations for extended courses of action that take place over varying amounts of time and modelling these things. Since she began working on this in her PhD, the research methods have vastly improved, and thanks to her research ‘we now have very good methods for constructing abstractions directly from a stream of data, which is pretty exciting.

‘Within academia the main obstacles honestly have to do with infrastructure - the amount of GPUs we have doesn’t compare with some of the industrial labs, and the fact that the AI scene is so active right  now poses some challenges because students do a lot of internships - they come to the lab for a few months, then go away then come again, so it’s hard to keep them focused on their thesis!’

AI is transforming every industry it touches, and Doina touches on the impacts of these progressions and technologies in healthcare as well as ‘areas that have a social impact’ such as education and the analysis of government data, and future cities.

Upon asking Doina her opinion of AI and the loss or creation of jobs, she said that it’s no different to any industrial revolution, and of course jobs will be lost and disappear, but new jobs will be created. When cars were first created, carraige drivers thought they would be out of work, and whilst ‘there aren’t a lot of carriage drivers around anymore, there are a lot of drivers! So we’re going to see the same thing with AI’. The jobs that will be created, says Doina, will be more interesting for people to do as well as more creative.

To hear what else we discussed and hear Doina’s tips on how to get started in AI, listen to the fifth episode of the Women in AI podcast here.