[LXAI@ICML 2019] Talking about Machine Learning, Diversity, and Inclusion with Pete Denman from Intel Labs

Pete Denman is a Senior User Experience Researcher at Intel Labs that is developing exploratory prototypes and working with developers, engineers, and ethnographic researchers to create future technologies. He also worked there in the Stephen Hawking project and has interesting stories about the new types of design and ML solutions that need to be created to enhance accessibility for people in a seated position, limited mobility, and that daily interact with computers. This is the resulting conversation between Pete Denman and Omar U. Florez, Diversity & Inclusion chair at LatinX in AI.

Hi Pete, welcome to LatinX in AI. How are you today?
Not bad Omar, thanks for having me.

We worked in the same team in the past. Can you tell us a little more about you for people that didn’t meet you yet?
Sure. Something that you may not see because I am very close to the camera is that I’m in a wheelchair, I broke my neck when I was 20 and I have been in a wheelchair for the last 30 years. That means that I have limited use of my arms and I cannot walk at all. I’m also dyslexic and that can be more like a challenge than a physical disability that I have. As far as my career goes, I have worked in pretty exciting projects here in the labs at Intel.

What do you know about Latin America, which country would you like to visit?
There are a lot of countries. I would like to visit Ecuador. Colombia looks beautiful as well, I have a friend that has a house in Providencia Island and it is amazingly beautiful. Chile looks nice too as it has such a diverse ecosystem going from mountains to the ocean so quickly.

Speaking of diversity and inclusion, how do you help to connect the areas of disabilities, learning, and education with technology?
I have been working in different areas. I was one of the designers that designed the interface for Stephen Hawkins. I am also working now as the leader in the International Dyslexia Association bringing people with those problems and showing to other people with these problems that they can achieve goals while working in big corporations.

What are the main challenges that you see when creating innovation for people with disabilities?
The only consistent thing when working solutions for people with disabilities is the inconsistency. When you are solving a problem for an individual, a lot of times you are solving a really direct problem for somebody. But other times you are trying to do something that is inclusive for many people as possible. That boils down to the universal design concept that is almost the opposite of individual help. An example of that may just be an electric door, an individual need would be putting the button in the exact right spot so somebody could use the door since this person may not be able to use the door or push the door.

In your vision, how can Machine Learning and AI help us to improve accessibility and inclusion?
As I was just talking about, you may want to create an algorithm following a holistic approach, but that AI may be targeted to an individual and work in a particular way that may be specific to an individual user. That’s what makes it very contextual and dynamic, just blind-coding for a universal design may lead us to miss a lot of people that could be in the edge cases. Well, AI will allow us to hit more of those edge cases while defining one broad solution with several complex adaptations.

At the same time, we don’t have lots of data to train our models for those edge cases…
Correct, but if you are trying to do something for an individual person, you have to collect data for that individual. If you are trying to do design metrics for a type of disability, then you may be able to have large groups of data. But for an individual person, you will have a problem of too little data. You are either going to have users participating in the design process or make broad guesses about how they are going to use it and do iteration and testing.

Maybe that is an important call for ML researchers to create new learning algorithms that learn with small data.
Yes, I totally agree.

LatinX in AI is a diversity workshop at ICML that creates research opportunities for LatinX talent around the world, in your opinion what is the connection between diversity and accessibility?
That is a really big question. Right now disability is not seen as a diversity. In fact, I am not considered diverse. But you would say the struggles and challenges that I see on a daily basis are substantial. I don’t really know all the challenges of somebody of color or sexual orientation, but I know the challenges that I face, and I feel like they should be indeed in that spectrum.

What is your message to the next generation of young researchers in ML that come from Latin America to ICML 2019?
You know, there are so many challenges when you are trying to get your career started. Just find your passion, whatever that happens to be. If you can, try not to get caught up in just getting a paycheck, try to do things that you are thrilled and passioned about because that is going to be ultimately the thing that is going keep you surviving in your career over a long period of time.

Source: https://medium.com/latinxinai/lxai-icml-2019-talking-about-machine-learning-diversity-and-inclusion-with-pete-denman-from-184e29be74e2

Leave a Reply

Skip to content