Ericsson recently held a workshop on Artificial Intelligence (AI) for Good. From the Terminator and the Matrix in pop culture, to facial recognition and ethical issues, AI has always been a hot topic – and it’s had a history of negative press.
In fact, AI is a force for good! How has AI actually helped in the fight against coronavirus — the crisis that’s changed all our lives? And, how is it helping in other areas, for example, preventing climate change?
Dr Aleks Krotoski, Host of Ericsson UnBoxed Office Social Series, said that they will be looking at the future of communication technologies. They will also look at AI, facial recognition etc., along with the ethical issues. AI has always been a hot topic. Well, what about AI for good? Or, AI for super good! We will look at how AI is helping us in the fight against coronavirus. How is it helping us in climate change?
White House call to action vs. Covid-19
Paul McLachlan, AI Research Leader, Ericsson Research, said that Covid-19 has been keeping America on the tightrope. The Ericsson R&D team has a technology for good program. “Within 24 hours of the White House posting a challenge for Covid-19, we sent a call for volunteers. We were expecting 20 or 30 volunteers, but we got over 400 volunteers. There were scientists, people with incredible communications skills, etc.”
For the White House call to action to battle Covid-19, Ericsson partnered with the National Institutes of Health, Georgetown University and the White House Office of Science and Technology Policy on their open research dataset challenge Covid-19. This challenge has 10 open tasks ranging from understanding Covid-19 risk factors to finding treatment protocols.
On the Keep America Running – Ericsson web page, today, we’re facing a new challenge, unlike anything we’ve seen or done before. We are in uncharted territory, as businesses, industries, and societies are grinding to a halt. For all Americans, network connectivity has suddenly become profoundly important. As the heroes on the frontlines are working tirelessly to keep people healthy, safe, and fed, Ericsson is honoured to serve them by keeping society connected. Everyone brings their individual strengths. We bring the networks. It is time to work together. Keep America running!
There was the latest and greatest in data science to understand all of this Covid-19 happening around us. There were issues to go through, from ethics, to genome sequencing. With this rich talent, we were able to deliver back to the White House. It was an incredible experience. There was data retrieving and sifting. We started the challenge where the White House had gathered many papers on Covid-19.
You have such a wealth of knowledge now. Using the data teams that we had, we built brand new models. We also built information retrieval models. We needed to find out what is the most relevant information, which involves a different type of data science. We have tried to summarize the data. AI for good is being facilitated right now. It goes beyond Big Data.
Business + sustainability
Cecilia Nyström, Program Manager, Data and Analytics, Ericsson Business Area Managed Services, noted that we are working with AI researchers and collaborating with the different stakeholders. We are trying to determine how we can use our expertise in the telecom domain and AI domain.
How we can use this information to help the society in the battle against Covid-19. We have also learnt about AI. There is a case to rethink cases, and what the AI algorithm is all about.
She added: “We also learnt about climate change. Sustainability is really a great topic. We can do AI to increase the efficiency of the mobile networks. If you have the opportunity to reduce the energy consumption, you can also reduce costs. The combination between business and sustainability is really powerful. The data volumes in the mobile networks are really growing right now. It also means that new opportunities are also being enabled by all of this. However, it can also have a negative impact on the energy consumption.
“We have worked a really long time to increase the energy efficiency of the networks. AI now acts as an additional tool that builds in more intelligence into the networks. One example you can use AI for in the mobile networks is to predict the traffic and use your resources. This will end up saving a lot of energy as well.
Jörgen Gustafsson, Sector Manager, AI Infrastructure, Ericsson Research, noted that we are supporting with insights, as we are also looking to flatten the curve, and protect the privacy, which is very critical. The telecom network has produced lot of data. We take this aggregated data and use advanced AI and squeeze out insights.
Ethics of using AI systems
Now, organizations have principles they adhere to, as far as using AI systems go. Cecilia Nyström said: “We are working on the EU guidelines. We are building the infrastructure that is critical for the society. Questions like security and robustness are always very important. AI also brings some additional challenges.
Paul McLachlan added that Ericsson has taken ethics very seriously. We are very mindful of ethics. We are also looking at how do we understand the models that are coming in. Not only are we pushing our best practices, we are also doing developments and further improve the methods. We are interested in data ethics and data governance. We are extremely sensitive as well. We take that trust and responsibility.
Next, there is the question about having robustness in AI. How can we use AI in network security? Jörgen Gustafsson said: “We are extending robustness into the AI domain. We can trust the network, and it is just as robust as the other networks. We are starting from a very robust platform when it comes to security.
As for the role in sustainable development, Cecilia Nyström said: “We can play a very active role. We adhere to many UN goals. Besides customers, we also have a big responsibility when it comes to our own operations. We also need to reduce our own carbon.”
Why have biased AI?
Another area that Ericsson is looking at is: how would you ensure that AI is not biased with anything?
Paul McLachlan said: “As we are building systems, we need to ensure that the systems themselves are not biased. We also need to have people who are not biased. We have a diverse set of people working on the AI teams. They work hand-in-hand to build models.
“What are our presuppositions that might be? We get a diverse set of perspectives to work with and talk to. Once a model has been built, we also take explain-ability very seriously. Every model ought to be explainable. If we have a feature, that looks suspect or otherwise, we look at it, and understand why it is behaving in this way. We look at it quantitatively and qualitatively. We also look at it from a social perspective, and very deep and technical perspective.”
Gustafsson added that when we create algorithms, we ensure they are not biased. Are they having any gender bias? We verify that as well. We are humans after all!
Using AI aspect of the Internet
Next comes the issue of looking at the AI aspect of the Internet. Jörgen Gustafsson said that when it comes to democratization, we ensure how to train the data, and whether there is any bias, etc. The algorithm will help the system to learn.
Paul McLachlan added that all voices are being heard. There is a rich network in the USA. The power connecting everyone is democratic. That is a fundamental shift: being able to share your voice. As we bring in AI, and there are 5G networks too, it can get a little more complex to operate a network.
There are factors like geography and distance as well. For example, there may be a conversation going on between San Francisco and somewhere in Sweden. There should be a way to manage the network complexity. Ericsson is making sure that the networks are always up and available. There is AI in the loop. We need to manage the network complexity.
AI can also be used for network bandwidth allocation. Cecilia Nyström felt that the complexity of the network automatically increases. We are using these types of technologies to manage intelligently.
Open sourcing is a big part to help a project to grow. Is there any plan for open sourcing Ericsson’s work? McLachlan said: “I am an advocate of open source code and network. However, we do operate critical infrastructure. We have to ensure that we are operating with lot of security and trust. We are very proud of our accomplishments.”
Democratization of AI
Further to the democratization of AI, perhaps, it will become a commodity. Gustafsson added that it is very important to think of the world when we talk about AI. Covid-19 is a good example of what we are trying to do for the good of everyone. Once we can scale it up, this will benefit everyone.
Nyström agreed, saying that you can also learn the language, and put it to use. There are lot of open courses available online. We can get people to be digitally literate.
As for privacy in AI, Gustafsson noted that there is the Covid-19 where there is non-sensitive data. We are also trying to understand how the network is performing. If we are not close to people, the virus will not spread. Authorities can give guidance about how to flatten the curve. We can take data that is almost useless, and suddenly make it very useful.
Full AI for 5G and future networks
Finally, when can we expect full AI for 5G and future networks? Jörgen Gustafsson said that as AI becomes fundamental, it will be an integrated part of the network. We need AI to be able to develop the networks. If you are using 5G networks anywhere, there are one or two AI elements that are also making it work.
Cecilia Nyström added that besides managing and developing, we will more and more get into optimizing. Maybe, you can also get a better network performance.
Paul McLachlan said: “A 5G AI network will be used by the consumers in the future. There could be XR/VR, so that people can interact and travel. We are also thinking about the Internet of Senses. We are looking at taste, touch, smell, etc. as well. The combination of XR/VR with AI will make it a very powerful technology.”