We thank Silvie Spreeuwenberg, founding director of LibRT and Program Manager at Simacan, for taking part in the AI for Sustainable Development Initiative. We had a great time talking with her about how the AI world is shaping up when it comes to solving global challenges using AI, and what all things are to be kept in mind. In our discussion with Silvie, she shared several insights, including:
- Digitalization as the first step for achieving AI
- Opportunities in AI on urban issues
- Importance of Explainability in AI
What Silvie Says…
In any operation “the show must go-on”, therefore, (these) transitions take time. Digitalization is always the first step and paves the path for AI technology.
– Silvie Spreeuwenberg
My company, Simacan, will support this transition by connecting all stakeholders in a supply chain. Our mission is to support the transition to autonomous vehicles and, with the use of AI, autonomous shipments.
– Silvie Spreeuwenberg
..Developers should pay more attention to available techniques to open the black boxes and understand how these techniques help them to create more reliable and less biased automated decision-makers.
– Silvie Spreeuwenberg
So, let’s dive right in!
How did Silvie start in AI
Chayan: Hi Silvie! Great to have you here. You have been working in AI for more than a decade now. We’d like to know that how did your journey start and how has it been throughout? Where are you heading now?
Silvie: Hi Chayan! I started with AI in the ‘Second AI Wave’: the time of expert systems in the eighties and early nineties. I was triggered by a philosophy class on the exam topic ‘Man-Machine Interaction’. The question of whether machines could complement humans in reasoning, and how human reasoning differs from mechanical mechanisms, interested me. Then, I came across a study that combined mathematics, biology, language, psychology, and philosophy named Artificial Intelligence and I joined for the simple reason that I liked all disciplines and could defer making a choice.
AI industry on Global Issues
Chayan: That’s a pretty interesting story, Silvie! So, a lot of research is being done in the use of AI for Sustainable Development. How is the focus of the Industry shifting towards providing solutions for these global challenges like Poverty, Social Inequality, Climate Change, Clean Water, etc.?
Silvie: These are important big challenges and the industry is shifting towards realizing these goals in small steps. In any operation “the show must go-on”, therefore, these transitions take time. Digitalization is always the first step and paves the path for AI technology. AI simply is able to optimize a complex process with many variables. Also, it makes complex knowledge accessible to a broader group.
Role of Simacan in Sustainable Development
Chayan: Absolutely, Digitalization would be the key step. On to the next question. How big a role can AI have in achieving Sustainable Development of the world? Are you looking to work on these global challenges?
Silvie: It is through efficiency gains and knowledge sharing that AI contributes to sustainability goals such as a CO2 emission reduction or wealth distribution.
AI helps people worldwide to do things that they could not have done on their own. Take, for example, the challenges in transport and supply of retail goods to cities. It is expected that 60% of the world population lives in cities in 2030. How to supply all these people with the goods they need while living standards increase and without a decreased quality of life? Cities are responsible for 70% of global emissions. City councils worldwide have no choice but to mandate transports in cities to become electric, introduce parcel boxes or use delivery droids. Shippers and carriers, reluctant to change, will use dynamic re-routing, real-time travel lights, and load pooling to operate with minimal disruptions. In either scenario, there will be lots of data from different stakeholders combined to optimize the transport puzzle on the day of operation and across supply chain borders.
My company, Simacan, will support this transition by connecting all stakeholders in a supply chain. Our mission is to support the transition to autonomous vehicles and, with the use of AI, autonomous shipments.
Thoughts on explainability in AI
Chayan: That’s great. Simacan could really be a game changer in the Supply Chain AI domain. So Silvie, You mention that explainability and transparency are crucial when it comes to AI. When it comes to AI at scale for global issues, do you think certain policies and regulations need to be set up to assess and control what these AI systems are understanding and predicting under the hood?
Explainable systems have a higher return on investment and therefore, should sell itself, without special regulations and policies. What is needed is awareness of decision-makers that explainable systems are also more trusted, easier to adjust, and more ethical. Also, developers should pay more attention to available techniques to open the black boxes and understand how these techniques help them to create more reliable and less biased automated decision-makers.
Tackling energy usage of AI
Chayan: Indeed those black boxes need to be cleared up! Another crucial point I wanted to bring up is that keeping the high-end AI systems running would require massive computational resources which will only be available through large computing centers, which in turn will result in a very high energy requirement and carbon footprint. Do you think this will inhibit the impact of AI on areas such as Clean Energy? How can we tackle this?
Silvie: This is a major concern and we may need to go back to times where we had to think very carefully about processing usage and volumes of data storage. Moore’s law – processing power doubles every two years while cost halved – made us focus on functionality. Processing power and storage capacity seemed endless. However, as with any other never-ending story, at some point in time, it will become a fairy tale. Due to my optimism about human intelligence and its creativity, I believe we will find ways that reduce the need for computational resources or its footprint.
Negative Impacts of AI
Chayan: I hope we do make a breakthrough in that space! Okay, so similarly, AI can have many long-term and short-term impacts in several areas. These impacts can be negative as well. What negative impacts can AI have? Do you think there is a need to systematically assessing the extent to which AI might impact all aspects of sustainable development?
Silvie: Any technology can be used for the good and for the bad. The question is if our existing framework of ethics, to protect humans from harm, is good enough to prevent harm from the negative impacts AI may have. I conclude that existing frameworks take care of most issues. However, AI may intervene with the fundamental rights underlying our democracy such as the ability to learn, freedom of choice, and equality. Governments and citizens are right to mandate regulations when these values are at stake. But, in the end, anyone creating and using AI technology has a huge responsibility to use AI in an ethical way. In my opinion, transparency and explainability contribute to that responsibility.
Perhaps a link to the book I reference would be helpful:
AIX: Artificial Intelligence needs eXplanation or my personal website www.silviespreeuwenberg.com and company website www.simacan.com.
Silvie’s Book AIX: Artificial Intelligence needs eXplanation
In my book on explainable AI, I list ‘being ethical‘ as one of the drivers to invest in explainability. I also research if and how AI differs from other technologies that may have a negative impact on society.
Chayan: That is awesome, Silvie! It was great having you on AI Time Journal and getting to know your thoughts and beliefs.
Okay! So that was Silvie Spreeuwenberg on AI for Sustainable Development. I think her thoughts on AI for transport, delivery and Supply Chain automation were quite on point. And she does focus a lot on explainabilty in AI, which is a matter of real concern when you want to understand what exactly your Machine Learning model is learning about and focusing on which parameters of the data.
I hope you liked this episode of the AI Interview series in the AI for Sustainable Development initiative by AI Time Journal. There are some very interesting talks lined up. Follow us and stay tuned!
