Natalia Efremova, a Research Fellow in Marketing and Reputation at Oxford’s Saïd Business School, discusses her work using machine learning and biometrics to address longstanding marketing problems and transform the consumer experience.

How did you come to work in artificial intelligence (AI) and specialise in neural networks?

I don’t think there is anything more exciting than AI, although when I first started working in this area back in 2002 it wasn’t nearly as popular and widespread as it has become.

I have worked in a number of AI fields, including natural language processing and computer vision. My main work now focuses on applying neural networks and deep learning to marketing problems, using computer vision tools such as face recognition, emotion recognition and biometrics to support development in consumer marketing.

How can AI be used to improve consumer marketing?

Whether they are targeted online, over the phone, through the post, or face to face, most people find marketing incredibly annoying. Better understanding people’s consumer habits helps us to make the experience more engaging and enjoyable, so you really do see what is relevant.

The key to great marketing is understanding consumer behaviour, and the people who buy your products. Neural networks-based applications can help us to understand what affects people and influences their buying decisions. 

How does your research support these goals?

I am working on a project that uses neural networks and biometrics to understand the attention mechanisms that underline people’s shopping behaviour. 

Online shopping is a key consumer environment now, but a lot of time and money can be wasted on online marketing. People tend to scroll through pages online so quickly that it can be hard to track what they are actually looking at and target shoppers accurately. AI allows us to track their gaze and see where their eyes fixate and what they are browsing online on their phones. What do they look at and what attracts their attention when they are browsing?

Natalia Efremova Natalia Efremova

Natalia Efremova

What are the biggest public misconceptions about AI?

I believe that people really think that AI is this ‘magic black box’ that can do anything on its own, which just isn’t true. The truth is, there are a number of different forms of AI, which are tools that can be applied to different situations but are dependent on and driven by data. People often say that algorithms are biased, but that’s not the case. The data that it is trained on is not representative, so the result is flawed. 

I think people tend to misuse AI technology. In my opinion it is best used on a large scale for social good, particularly for supporting sustainability goals. In addition to my work on marketing and AI, I try to collaborate with peers in Oxford’s School of Geography and the Environment and Department of Computer Science on projects that use AI to improve the environment, understand climate change and prevent species extinction. For example, we are developing software than can make freshwater source predictions, shedding light on areas that are at risk of droughts and episodes of extreme weather. 

Do you think AI has more potential for good or bad objectives?

In general, the industry is very balanced, so there is lots to be cheerful about. But there will always be people with bad intentions in this world, and every single piece of academic research can be taken out of context and applied to support bad goals. 

We have to stay optimistic though and continue to do our best to support AI for good, because there are plenty of people fighting the good fight.

What excites and inspires you most about AI’s potential for societal good?

There are fields that we have to work on for the betterment of mankind, like climate change and sustainability goals. 

I am personally most excited about how it can help the fight against climate change because it is so, so important.

People don’t want to hear this, but they need to. The trends are really scary and we need to start working on how we will live on this planet when things change and the temperature rises, because it is going to happen, and soon.

What are your biggest concerns about the field?

There is too much hype about AI, and as a result people just do not understand what it is.

I see many students now writing that they work in AI on their CVs, but their understanding of the field is minimal. We need to do more to support public understanding of AI, and to make computer science in general more accessible, so that there is less fear around the technology.

Did you always want to work in science?

I actually did my master’s degree in natural language processing at home but since then I have gradually developed an interest in artificial intelligence and computer vision. I have worked for a few years as a machine learning researcher in computer vision in a startup and studied at the Executive MBA programme here in the Saïd Business School at the same time. 

I am currently doing a postdoctoral degree with Andrew Stephen, Associate Dean of Research, L’Oréal Professor of Marketing and Director of the Oxford Future of Marketing Initiative.

I had been working as a data scientist for about seven years in many countries across the world when I decided to pursue a PhD in computer science in the University of Kyoto, Japan. I hoped that the qualification would help me to get to a fairer place in my career and that people would start to respect me more as a professional – which it did. 

Can you elaborate on what you mean by a ‘fairer place’?

As a woman working in science I have always felt a little isolated, particularly at the beginning of my career in Russia. And then, when I went to Kyoto University to do my PhD, I was the only woman ever to do the course at the time.

Things are changing, but there are still not nearly enough women working in STEM and it can be hard to connect if you are.

There is increasing weight on women’s voice in AI. In some areas of academia there is a quota for female researchers. Even though these measures are forced, and women and minorities are encouraged to apply, I think it is very encouraging. But I don’t think there would be a natural drive to improve diversity within sciences without them.

I am naturally quite selective about where I apply. On some level I feel that my application will not be considered as seriously as my male peers. I try not to take it personally, though – statistically, more men apply for positions in AI and computer science than women. I hope that changes, but for now at least the odds are naturally in their favour. 

What do you think can be done to make STEM environments more inclusive?

Networking opportunities and connecting to academic societies definitely help. I am an active member of the women in machine learning community and recently presented at the NeurIPS  conference, which was fantastic for networking and potential collaboration. 

I think that Oxford’s women in machine learning community could be much stronger. I’d love to see some kind of mailing list that makes it easier to collaborate with people in other departments. 

I try to connect with as many people as possible, but it is not that easy. If anyone is working on attention tracking on mobile devices, please get in touch!

What advice would you give to a woman wanting to work in AI?

‘Please come, there are many opportunities!’ Applying somewhere like Oxford can feel daunting and overwhelming. But sometimes you actually can do what you want, you just need to overcome your fear that you can’t. I would never have thought that I would one day be working in AI at Oxford. If I can, you can.

What are you most proud of in your career so far?

Some regions of our planet are so poor that there is often no funding available for crop monitoring and hardware for precision agriculture, such as irrigation systems or soil moisture sensors. 

I am really proud of my project with the European Space Agency. I work with them on a not-for-profit project predicting crop stress and soil moisture in the developing world, which helped farmers to increase crop yields and reduce waste of scarce resources. We use machine learning to analyse earth observation data to assess soil moisture content and help farmers in Africa to maintain their crops better, improve their irrigation, and know what to plant and where.