This blog continues the summary for our AITAB meeting in November. Given the diverse group of voices, we were able to cover a lot of ground in the area of AI for good. In the first part I introduced the 3 main trends of AI for good: democratization of AI skills, green AI, and AI justice. In this blog, we cover examples of AI for good in the industry, academia, and a global perspective from Eastern Europe. Our board members spoke from experience and also listed some great resources to anyone interested in getting deeper into the field.
AI for Good in the Industry
Davi: Another way AI and machine learning are helping in sustainability, is by improving companies’ consumption of non-renewables. For example, one of the largest expenses of a cruise line company is fuel consumption. Mega ships require untold amounts of fuel to move them across oceans around the globe. And, in maritime, the exact same route may require differing amounts of fuel due to the many variables that impact fuel consumption such as the weight of the ship, seasonal and unexpected currents, and different weather patterns.
AI and machine learning have expanded the capacity to calculate, with never-seen-before precision, the amounts of fuel needed for such mega-ships to safely complete each of their routes in real-time. This newfound capability is not only good for these companies’ bottom lines but also helps them preserve the environment by diminishing emissions.
Elias: That’s a great point. A recent study by PricewaterhouseCoopers estimates that AI applications in transportation can reduce greenhouse emissions by as much as 1.5% globally so this is definitely an important trend to track.
A Report from Eastern Europe
Frantisek : I tried to investigate and revise my knowledge in the three areas Elias proposed. Regarding the first topic, democratization of AI skills, I think from the perspective of Prague and the Czech Republic, we are at the crossroads between Eastern and Western Europe. There are initiatives that focus on AI education and popularization and issues related to that. I would like to point out a specific Prague AI initiative corporation of different academic and private companies as well.
This kind of initiative is more technological, and they are just beginning to grasp the idea that they need some ethical additions like philosophers. Yet, they are not inviting theologians to the table. I guess we need to prove our value before we will be “invited to the club”.
With that said, Prague AI wants to transform the city into a European center for AI. They have good resources for that, both human and institutional support. So, I wouldn’t be surprised if they achieve this goal, and I wish them all the best. My research group aims at connecting with them too. But we need first to establish ourselves a bit better within the context of our university.
On another front, we established contact recently with a Ukrainian Catholic University which aims at opening an interdisciplinary program for technology and theology. However, we do not know yet, how far they are with this plan. We intend to learn more since I am in process of scheduling an in-person meeting with the dean of their Theological Faculty. It was not yet possible due to the pandemic. We are very much interested in this cooperation.
We also aspire to establish a conversation with the Dicastery for Integral Human Development and other Vatican-related organizations in Rome where issues of new technologies and AI receive great attention especially in relation to ethics and Catholic Social Teaching. In December 2021 one of my team members went to Rome to start conversations leading towards that aim.
In summary, here in Central and Eastern Europe democratization of AI is more focused on education and popularization. People are getting acquainted with the issue.
Regarding sustainable AI, we are following the footprints of the European Commission. One of the European commissioners that formed this agenda is from the Czech Republic. And maybe because of that, the European Commission sponsored a big conference in September which was in large part focused on green AI. The contribution about the role of AI in processing plastic materials was especially interesting because it has a great potential for green AI implementation.
The European Commission introduced a plan for the third decade of this century. It’s called the Digital Decade. It includes objectives like digitalization of public buildings, digital economics, and the growth of digital literacy among citizens with large support for the field of AI.
In Europe, AI justice is a big issue. There is hope and a lot of potential in AI to contribute towards the effectiveness and quality of judicial procedures. Yet there is an equivalent concern about the fundamental rights of individuals. I’m not very well acquainted with these issues, but it is an important topic here in Europe.
AI for Good In Academia
Scott: I’m a professor of physics at Belmont University. I started working with machine learning around 2013 /2014 with respect to developing audio signal processing applications. I managed to get into an AI Ethics grant in 2017 and went to Oxford for a couple of summers.
My background a long time ago was astrophysics, but recently, I eventually became focused on machine learning. I split my time between doing technical work and also doing philosophical and ethical thinking. I recently taught a general education undergraduate class that integrated machine learning and ethics. We talked about how machine learning and algorithms work while also discussing various ethical principles and players in the field.
Then the university requested I teach a more advanced course more focused on coding for upper-level students. This fall I’m teaching AI deep learning and ethics, and it’s kicking my butt because I am writing a lot of the lessons from scratch. One of the things I’m doing in this course is integrating a lot with things from the open-source community and the public free machine learning and deep learning education. There’s Google, Facebook, and then there’s everybody else.
So I’ve taken a lot of classes online. I’m pretty involved with the Fast AI community of developers, and through their ancillary groups like Hugging Face ,for example. It’s a startup but also a community. This makes me think in terms of democratization, in addition to proliferation around the world, there’s also a proliferation with everybody that’s not at one of these big tech firms as far as disseminating education.
Democratization of AI and Open Source
I think a couple of big things that come to mind are open source communities that are doing their own work like Luther AI. They released their own GPT model that they trained. It’s a sort of grassroots community group that is loosely affiliated but managed to pull this off through people donating their time and expertise.
One of the things they teach in Fast AI is a lot about transfer learning. Instead of training a giant model from scratch, we’re taking an existing model and fine-tuning it. That relates to sustainability as well. There are ecological concerns about the power consumption needed to train language models. An example would be Megatron-Turing Natural Language Generation (MT-NLP) from Microsoft, a gigantic language model.
With transfer learning, we can start with an initialization of a model that doesn’t require much power. This allows people all over the globe to run them with little computational resources. The idea is to take Ivory Tower’s deep learning research and apply it to other things. Of course one of the questions people think about is what are we inheriting when we grab a big model and then fine-tune it. Yet, nobody really knows how much of that late structure stays intact after the fine-tuning.
It’s an interesting and accessible area. Considering how many people, myself included, post free content online for education. You can take free courses, free blog posts for learning about machine learning, developing tools and ethics as well. The open-source movement is a nice microcosm of democratization of content that relates both AI ethics and sustainable AI.
Elias: Thank you, Scott. I want to seize on that to make a point. Open source in the tech world is a great example of the mustard seed technology idea. It starts through grassroots efforts where many donate their time to create amazing things. That is the part I think technology culture is teaching theology to us by actualizing the gift economy. In the real world we live in we pay for companies and they focus on profit. It is highly transactional and calculating. Here you have an alternative economy where an army of volunteers are creating things for free and inviting anyone to take it as needed. They build it simply for the pleasure of building it. It’s a great example.
Scott: I’m also starting to do some work on algorithmic auditing, and I just found this kid from a group called data science for social good. Other people may find it interesting as I do.
