On 15-16 November, the women’s network Women@Telekom sponsored by Deutsche Telekom threw a hackathon at hubraum. Participants came from all over Germany to take part in three challenges: hacking the future of mobility, #metoo and countering subconscious bias in recruitment, all competing for a cash prize. But the event wasn’t just about competing, but learning. Preliminary talks focused on the severe lack of women in AI and what problems this might create for the industry.

Hubraum spoke to three women present at the event to delve into why this might be the case. We spoke to Kenza Ait Si Abbou, a Senior Manager of Robotics and AI at Deutsche Telekom IT (and who also heads the umbrella organization of women’s networks at Deutsche Telekom); Claudia Pohlink, who leads the Intelligence Group at Telekom Innovation Laboratories and Vidya Munde-Müller, the founder of Givetastic.org, a non-profit aiming to improve the charity sector via the power of AI.

How did you all get interested in AI?

Kenza: 15 years ago while studying telecommunications engineering, I got interested in machine learning so I wrote my thesis on clustering algorithms. I wanted to delve deeper into AI but then life happened. I moved for work to Barcelona, Berlin, China and then back to Berlin. After seven years working at Deutsche Telekom, I decided to get back into AI. After asking around, I found a project in Deutsche Telekom’s internal IT department, where they were putting together a new team to build our AI platform — this was a year ago.

Vidya: I got into AI in a more roundabout way. I was living in Silicon Valley for a few months on a job rotation, working as a business partner manager and I needed to find startups working in AI based in the Valley. I’d done one course on machine learning so I had the basics. I went to so many conferences, I got to know so many people from Google, Apple, Twitter and had these great conversations and I felt like I picked up so much in such a short time. I was completely taken by AI and thought “Wow, this technology is going to change our lives.”

Eventually I came up with an idea for how to use AI for social good and this is the idea I’m pursuing still. I thought: why not help increase the number of volunteers or donations using AI? After all, we’re all always consuming news via the internet. But 78-80% of donations take place via analogue means. Clearly, there’s still space to leverage the power of the internet for social good. So my project, Givetastic does the following: Maybe you’ll be reading a report about animal poaching – the prototype shows you charity recommendations based on the article you’re reading. We want to go further with this and show volunteer slots related to what you’re reading to. We’d also like to make this recommendations location-specific, so if you’re in Heidelberg, you can see voluntary opportunities close to there.

Claudia: I’ve been working for T-Labs for around seven years and I started working in data analytics here. Eventually I left the labs and started working in the chief data office.
Two months ago, I decided to move to T-Labs again and to take over the lead of our machine learning team. And now I head up a team of ten people. Well, ten men — right now, they’re all men but I’m hiring a woman. It’s not always that easy, I can tell you. There’s a lot of experts in the team who have a background in physics or engineering or have a PhD so they’re all smart and experienced and then there’s me, this extroverted woman. But I love challenges, so it’s fun.

According to Wired, just 12% of employees in machine learning research companies are female and this number gets even worse at Google where only 10% of their AI workers are women. Why do you think so few women work in AI?

Vidya: You said 10-12% and I have even heard 8% for machine learning. For the tech population in general, only 20% of workers are women, so it’s not a problem that’s limited to AI, I think. The bias starts from an early age, long before women are old enough to work.

Kenza: Right, ask yourself how many girls end up studying sciences, how many girls end up doing STEM? Engineering? Very few.

Vidya: I think it might also be because there are so few women in AI, which means there are few role models. I’m so proud of the women sitting round this table – Claudia, Kenza – because they have reached a senior level. And I hope that they get promoted to even more senior roles. I would love for there to be more women at the top because women will take heart. When they see someone else did it, maybe they’d begin to think about such a role being possible for them, too.

Claudia: I tell you what though, I don’t want a more senior role.

Vidya: (laughs) Well, you deserve it!

Claudia: It’s important to talk about role models, but we should also acknowledge reality: it can be hard to balance work and your personal life. I have a ten year old daughter and she’s always asking me “Why do you work such long hours? Look at the other mothers. They’re always there for their daughters, so please, can you be there too?” That can be difficult.

Kenza: The problem is very complex. If your work environment isn’t flexible and doesn’t allow you to pick up your child from nursery at 4, 5pm and then continue working after you’ve brought them home, then maybe you can’t do the job. At the moment I’m very happy where I am — I’m completely flexible.

Claudia: Same for me. I’m also flexible working at the level I’m at.

Kenza: I know our company is doing a lot. Our company has an incredible infrastructure. I don’t have to work from the office, I can work from home whenever I like, I come here whenever I like, I can work from anywhere, I can work anytime. My boss is happy as long as the work is done. So from the company side, Deutsche Telekom is offering us the best possibilities.

But challenging people’s biases means that changing society will always be about changing the way people think, not changing the structure of a company. Take one example: If your boss is fine with you working from home, that’s great. If you have a second boss and he’s not fine with flexible working and he doesn’t trust you, he might not allow you to work from home even though the company allows you to work from home.

Kenza Ait Si Abbou, Claudia Pohlink, Vidya Munde-Müller

From left to the right: Kenza Ait Si Abbou, Claudia Pohlink, Vidya Munde-Müller

Have you experienced challenges working as women in AI?

Kenza: Sure. In the past, when I was looking for new opportunities, I had an interview with one guy. He told me “Oh, your profile is just what we’re looking for, you’re perfect for the role.” A couple weeks later I found an email from HR saying sorry, you didn’t get the job. I phoned the guy to see what had happened and he said “We’ll talk about it in the future, but right now, I’m looking for someone who’s flexible.” So I think, ok, what do you mean by someone who is “flexible”? And he clearly thinks because I’m a mum, I’m unable to be flexible. This is the thing that’s very difficult to change.

Claudia: I haven’t really experienced sexism, quite the opposite: my new boss John Calian actually supported me in applying for my new role. I wasn’t self-confident enough to consider the role until he said “Claudia, please do apply for this job.” I thought I couldn’t do it, I was worried the men wouldn’t respect me. He said “Come on, let’s try this. Let’s treat it as an experiment.” He empowered me.

Vidya: I’ve also experienced sexism in the tech world: if you’re a project manager and you want things to be delivered on time, you can get characterized as a woman as bossy, as pushy.

Kenza: I have another story: 12-13 years ago, in Spain, I was applying for a job in electronics, the field I was working in back then, and I got to the interview and the man said “I can’t hire you, you’re too beautiful.” I kept a poker face. He went on, “Look, you’d be the only woman here. If you came into the office everyday, nobody’s going to do any work anymore.” I didn’t know what to say. I didn’t get the job and it was fine – I mean, it was such an awkward situation.

Do you have small practical tips for smaller companies – an AI startup, say – to make their environment more welcoming for women?

Claudia: I’m not sure this is a tip for women only but the best work in data science and AI doesn’t necessarily take place in an office. Sometimes you have an idea and then you dig deep into that topic, you’re working on that topic all night and you want to keep developing your idea. As such, you don’t necessarily need a desk or an office, since some of your best work will happen at home. But as a boss, you have to be open to that kind of model.

Vidya: There has to be some flexibility with working part time, jumping around.

Kenza: It would be great if you could bring your baby to work. I think these sort of things which make women’s lives easier haven’t been a priority in the past and now it’s becoming more important.

I also think for women, having a good team is really important. The moment you have a bad team, no matter how interesting the project you’re working on is, you’ll want to leave. Chemistry between people matters. And I think startups have an advantage here because when you hire people at a startup, you can have the whole team of six people in the room during the hiring process. The interviewee comes in, meets her future team and if everyone clicks, you’ve got her.

AI HACKATHON LADIES ONLY, hubraum, Deutsche Telekom AG, Berlin, 15.11.2018

Participants from all over Europa got together to challenge themselves and exchange knowledge

Why do you think it’s important that AI teams be more representative?

Claudia: AI is our future and AI will be everywhere. As such, AI should be for everyone. That means that there needs to be diverse teams behind the products. When you tell me that 10% of Google is female, that means to us that the projects will be male-dominated but projects should be neutral in every possible aspect. Which means the composition of the team itself should also be neutral.

Vidya: There is a huge diversity problem in AI. The problem is that if we don’t challenge this absence of women, then we’ll get faulty programming because everyone brings their biases to programming.

Think about the commercial facial recognition software that Microsoft and IBM pioneered – it didn’t work on darker skins, so imagine my skin or Kenza’s skin – the software wouldn’t pick up our faces. If we don’t make proactive changes right now, if there aren’t enough women or people of colour at the table contributing to data or algorithms, the products will be skewed.

Women In AI was founded to close the gender gap in AI — you can find more information about the organization here.