Eleonore Ferreyrol : Main Challenges of AI
So Eleonore, who are you and what do you do in Singapore?
I was working in Paris and I decided to move to Asia 6 years ago. I had job opportunity in Singapore so we decided to move here. I coordinated the setup of the regional treasury for a large organisation.
Singapore is such a thriving hub of activity and there are a lot of opportunities to seize! I thought “why not step out my comfort zone?” and I chose to leave the corporate world. I wanted to dive in the startups’ landscape, where innovation happens.
I joined several network and organised events to gather startups and investors from Australia and ASEAN.
Down the path, I joined a Singaporean RegTech startup where I led the sales & strategy activity.
Beyond that I am also a founding member of Live With AI, a think tank on Artificial Intelligence.
What is “Live with AI”?
We created Live with AI as an independent initiative during the France Singapore year of Innovation in 2018.
It is a think tank that gathers thought leaders, academics, startups and corporates. Artificial intelligence can have a positive impact on society as a whole and we want to show how. We drive several work streams and research projects to investigate these aspects.
Last year we published a report to propose 14 recommendations to leverage artificial intelligence in our daily lives, and how we can prepare ourselves.
A “Live with AI” delegation went to Viva Technology 2018 in Paris to launch this report, and it has been downloaded 3700 with zero advertising.
We address several use-cases in the white paper, like the need for continuous training models or how to move toward greater financial inclusion thanks to AI.
We hear everything and anything on AI. How do you define it?
Some say “AI could eventually have god-like capabilities”. Others say that “AI is a fundamental risk to the existence of human civilization”.
Let’s take a more standard definition. Artificial Intelligence refers to the theory and development of computer systems able to perform tasks normally requiring human intelligence such as speech recognition, problem-solving, learning and planning.
Indeed it is a very popular subject that is widely discussed in technology and business circles. What’s interesting is that it’s not a futuristic technology, it is already all around us. From Siri and drones to virtual assistants and software that translate or invest, or even algorithms used to predict our cultural interests on Netflix…it’s everywhere!
Where do you see applications of Artificial Intelligence?
Potential applications are huge and offer great opportunities.
In Singapore, AI has already taken a strong foothold in healthcare. As an example, a mobile application can help nurses to assess chronic wounds and presents a preliminary assessment. This can empower the nurses, to get insights faster to then provide a more accurate diagnosis. By saving time in their research they can also spend more time with their patients and taking care of them.
“Smart agriculture” programs have been developed to help farmers to monitor crops more effectively and make better predictions to then reduce the water and pesticides consumption.
The manufacturing sector strongly benefits from Artificial Intelligence. Studies show that unplanned downtime costs manufacturers an estimated $50 billion annually. Thanks to predictive maintenance using machine learning and neural networks, asset malfunction can be predicted before the assets break down, which reduce the costs related to the slowdown in production for example.
What are the main challenges of the adoption of Artificial Intelligence?
For a broad adoption of artificial intelligence, there are various challenges to address. These include building trust, by ensuring fairness and explainability of the outputs. We can not use a tool at a large scale for loans application for instance if there is no transparency on how we get the results and if the outputs can’t be explained. This black box problem is not new, and the more the model becomes complex the more it’s difficult to explain.
We need also to be careful of the historical data used to feed the model that may reflect the past unfairness in the system, like promoting a specific group of the population.
In this regards, the Monetary Authority of Singapore has released a set of guidelines, FEAT, Fairness Ethics Accountability and Transparency in the use of AI and data analytics in finance.
This highlights that even if we can delegate some of our tasks, we do need to have a human in the loop to assess the results from the AI systems.
Do you think gender diversity could help?
Recently, Amazon tested a system for automating the recruitment process. The input data where biased: males have been dominant in these technical roles over the last decades.
Men and women won’t necessarily think to ask the exact same questions of the data, because they may define problems and possibilities in quite different ways. That’s why it’s important to have diverse profiles who look into the data and the algorithms designed to use Artificial Intelligence.
And it’s not only about gender diversity but also ethnicity, race, language, skin color and age. Otherwise, this may lead to having a pigeonhole view of society.
Joy Adowaa Buolamwini is a Ghanaian-American computer scientist and digital activist based at the MIT Media Lab. She is working to compel organizations (such as IBM) to make facial recognition software more ethical and inclusive since she realized that facial recognition software can be bad at assessing darker male and female faces.
A final word?
One of our mandate with Live With AI is to gather a diverse community of thought leaders. We have various professional backgrounds, age range and nationalities. But we are still looking for more female experts in Artificial Intelligence and I am sure there are! So feel free to contact me if you would like to contribute and share more about your work.
Eleonore, thank you ! Linkedin: Eleonore Ferrerol
Get in touch with us @ womenfrenchtech at gmail dot com
In collaboration with Amel Rigneau & Sophie Guo
