Machine Learning Researcher Galina Alperovich explains AI and relates her experience as a woman in a male-dominated professional field.
It’s International Women’s Day. In observation, we’re taking a look at the male-dominated tech industry through the eyes of one of its female professionals, one of our top researchers, Galina Alperovich. Galina is a Machine Learning Researcher at Avast, and we asked her to share some of her experiences and thoughts on the topic.
Avast blog: Hi Galina, thank you for taking the time to talk to us about your experiences of working in the tech industry. Can you tell us a bit about what you do here at Avast?
Galina: I work in the Research and Development department, specifically on the applications of machine learning and artificial intelligence (AI) for security. I’ve been working on projects related to Device Identification and Anomaly Detection in network traffic. I’m also developing internal tools and systems for our machine learning pipelines. I have been a part of the Avast team for almost 2 years.
In recent years, the general perception of AI has often been confused by the term being overused – the term means completely different things to different people. AI is a very broad and rich field with many practical applications and research topics. In fact, everything that is even slightly related to machine intelligence is often labelled as AI (and that actually annoys a lot of ML researchers!)
Avast blog: What has been your experience of working in this field? Can you tell us about the challenge with diversity in tech?
Galina: AI is a STEM (Science, Technology, Engineering and Mathematics) discipline and there are historical contexts for gender bias in this area. In the past, societal and parental pressures dictated what a girl should or should not do in life – and this usually did not include pursuing a STEM career.
Today, if a woman chooses to step into a technical field, she can still face initial gender stereotypes and prejudices which could easily result in performance failures. Once employed, women are still not always free to get on with their job without perceptions from teams and management that she is not capable of doing the required work.
Even in companies with a good working culture, women can still be trapped in traditional environments where old stereotypes may surface. It’s important to recognize how this can lead to subjective decision-making, lost opportunities, and an increasingly toxic environment.
Avast blog: Do you think women face these stereotypes often?
Galina: Of course – while it’s unlikely that all these things will happen to one person, I have experienced some of these situations and my female friends have also shared stories of gender bias in their workplaces.
I have met many manifestations of the bias – most of them related to prejudices from men, although most are implicit as it’s rare that a man will explicitly say something degrading directly to you. It’s hard for me to say how much it influences my career because I am already “in the system” and don’t know what I miss because of gender. It is rare for somebody to speak about it publicly, but the patterns are clear that women are treated differently.
I studied math in high school, at two universities in two countries, and worked on four teams during my career, and I can say that I see gender bias almost everywhere. It’s often the case that I’m either the only woman or one of the few women on a team.
This is endemic in society and these prejudgements follow women throughout our careers. Our professional lives necessitate communicating with thousands of people, whether in-person or otherwise. The more people who hold on to these outdated prejudices, the harder it will be for any individual to walk down this road.
Another important cause of the AI gender bias is a lack of education in AI. Not all universities have such programs yet, and naturally, a lot of AI specialists are self-taught professionals from other STEM fields – most of whom are men. Significant achievements in Deep Learning (a sub-field in ML) have led to an enormous amount of tools and learning materials from top companies, professionals, and universities posted online and available to anyone. This is a good step in the AI “democratization”.
Avast blog: And how do you see this bias affect the work you do and the AI domain?
Galina: To speak broadly about AI, we should first consider different sub-domains of AI separately because I believe that the bias is not equally distributed among them.
For example, one of the fundamental AI tasks – creating an intelligent agent (i.e. an autonomous self-driving car, an industrial or human-like robot, etc.) requires many technical skills and involves various disciplines. To create a robot’s “body”, you need to know about hardware, sensors & signals, programming, physics, math, etc. To create a robot’s “brain” which is intelligent, you apply such techniques as machine learning, pattern recognition, planning, gaming, logical reasoning and more. These are each a different discipline of AI and each has a strong gender bias – I think it’s the intersection of all these disciplines that demonstrates a widespread structural bias in this particular robotics related field.
On the other hand, there is Data Science, another discipline related to AI, which is not about intelligent agents but intelligent data analysis, pattern recognition, predictions, and knowledge discovery. It is the closest AI domain to business as companies have many data records which should be analyzed to answer specific questions and help make optimal decisions. These techniques are widely used and implemented in various tools with excellent visualization. A whole new domain exists on top of these tasks called Business Intelligence. I believe that gender bias is the lowest in Data Science and Business Intelligence, even though it also requires technical skills, analytical thinking, and from time to time programming. Communication skills here are crucial because you need to have an idea about business processes, understand the particular problem, analyze data for it, formulate and present the results. In my experience, there are quite a lot of women on such teams and this shows that gender bias does not apply to all AI-related fields.
Machine Learning —that subdomain which is is more about intelligent computer programs which can learn and generalize from observed data. Machine Learning is concentrated on building the models and requires knowledge of statistical learning, programming, mathematics (especially optimization, algebra, and calculus), and overall data science techniques. I find it to be more technical than data science or business intelligence but in some sense, less technical than, for example, building a robot. In my experience, gender bias here tends to be somewhere between these two.
Of course, there are many other subdomains of AI, but the gender bias tendency I think is similar to other domains – the closer to technical and engineering layers, the more bias is found.
Avast blog: Do you have any thoughts on how the tech industry – and more widely, society as a whole – could help to overcome this challenge?
Galina: To be successful, everyone needs to understand the causes of the problem and contribute to tackling them.
Girls need to be strong and follow their dreams despite the stereotypes. This statement is true for everyone, however, it has been my experience in the past that sometimes a woman will need to work harder to prove something than a man in the same position. She often has to speak up louder, maybe even report inappropriate workplace interactions like disrespectful comments, and always be prepared to stand up for her rights.
The environment in which tech teams work is important. Teammates should respect each other and be able to evaluate work objectively despite gender or other differences. Organizations need to support diversity and work to establish a positively motivated, non-toxic working environment.
Society is not perfect. We are still in the process of building a society free of stereotypes and bias, and I don’t think it’s going to be solved soon. However, I see great changes in this area, and they have made me hopeful. I’m glad that at Avast, we do look to encourage women across all areas of the business and there is a strong culture of inclusion for everybody here.