Q&A with Google DeepMind’s Jane Wang

Jane Wang, a senior research scientist at Google’s DeepMind, is one of the UK’s rising stars in cutting-edge artificial intelligence (AI) research.

With a background in computational and cognitive neuroscience, complex systems and physics, Jane’s research has brought a fresh and multifaceted take on AI and its potential. She also writes some mean AI fiction on her blog.

We got in touch with Jane to find out more about her work and some of the most exciting research across the breadth of topics that make up machine learning (ML) and AI.

Simpleweb: What interests you most about the intersection of physics, neuroscience and AI?

Jane: I’ve always found myself intrigued at how different ideas in different fields can emerge independently from each other and, going from one field to another, you can really spot the connections and draw new ones that allow you to import new tools and perspectives to solve old problems.

Neuroscience in particular is very broad and encompasses many different scales (from the molecular all the way to the cognitive and even societal scales), so you really need to be able to shift among different fields and their respective languages.

SW: Which piece of research that you’ve been involved in are you most proud of and why?

Jane: Probably the meta-reinforcement learning work that’s come out over the past couple of years, because it really ties together my interests in AI and neuroscience and shows how the two can really feed into each other.

 

I’ve also always been interested in how humans learn, and it’s fun applying the insights gained from the study of human cognition to machine learning models, especially in reinforcement learning.

SW: What was the last ‘epiphany moment’ you experienced in your line of work?

Jane: Science is less about ‘epiphany moments’ as it is a series of what appear to be endless mysterious results, questions, setbacks, and incremental gains. When you can make a story out of all that messy work is when you finally publish, and looking back on it you can see the progress you’ve made, but in the moment it’s sometimes hard to feel like you’re getting anywhere.

The closest thing to an epiphany moment I’ve had is when I have a weird idea and think “hmm I wonder what would I happen if I tried this?” and it works! But it then takes months to figure out why.

SW: Which is the most exciting piece of research you’ve seen on AI?

Jane: This is a hard question… There are many aspects of AI I find fascinating, and there’s a ton of work being done in all of these. I really like work that aims to really try to understand why ML models work the way they do, like Jason Yosinski’s work, or this paper on deep learning by Dian Lei, Xiaoxiao Chen and Jianfei Zhao.

There are also a lot of great review articles being written that highlight the intersection of neuroscience and AI, like this one entitled ‘Towards an integration of deep learning and neuroscience‘. So it’s difficult to choose.

Big thanks to Jane for taking the time to answer our questions. You can find out more about her AI research at janexwang.com.