Video transcript: Collective Learning - Human-centered Distributed Artificial Intelligence
Transcript for the Collective Learning - Human-centered Distributed Artificial Intelligence video embedded on the Collective Learning - Human-centered Distributed Artificial Intelligence page.
[The screen shows the University of Leeds logo.]
[Music]
[The screen shows a bus driving down a street, a roundabout from above, and cars in traffic queues.]
[A voiceover says:] Social technical systems exhibit unprecedented complexity any imbalance can cause blackouts and traffic jams with catastrophic impact.
How can we combat these tragedies of the commons in the digital era? A sustainable development pathway for our digital society requires new means to self-regulate the consumption and production of physical resources such as energy.
[A thermostat, electricity pylon and field of solar panels is shown on screen.]
[A voiceover says:] In particular, power grids experience significant risks and uncertainties.
The penetration of electric vehicles and distributed energy resources, highly volatile production from renewables and an unpredictable demand during times such as pandemic lockdowns.
All these challenge such systems to adapt and prevent catastrophic blackouts.
The digital transformation promises new means to combat such uncertainties.
[The screen shows a person adjusting an electronic thermostat on a wall, a sat nav device on a car dashboard, then a person using a smartphone.]
[A voiceover says:] Smart interactive sensors in our homes, in our vehicles and in our phones provide revolutionary opportunities to collect fine grain data based on which system operators and power utilities can predict instabilities, and promptly respond.
Sensor data though, can often reveal sensitive personal information, from occupancy and lifestyle to mobility patterns and even what TV programs we watch.
[Dr Evangelos Pournaras, Associate Professor at the University of Leeds, appears on the screen.]
[Dr Evangelos Pournaras says:] Such personal data are often the basis for algorithmic discrimination and manipulative nudging, putting at risk our autonomy and democratic values on how we make decisions.
We envision a radically alternative human-centered approach based on the concept of collective learning.
[The EPOS logo appears on the screen.]
[A voiceover says:] The EPOS project has applied collective learning to power grids to prevent blackouts, transport systems to minimize traffic congestion and load balancing of bike sharing infrastructure and stations.
[The screen shows two maps showing congested and optimised traffic patterns through moving dots. The screen then shows a row of rental bicycles in a street and then a scrolling page of computer code.]
[Dr Evangelos Pournaras says:] So this is the open-source code for collective learning. It's accessible for the broader community.
[Dr Evangelous Pournaras appears on screen, sitting at a computer. The EPOS logo and maps can be seen on his computer monitor. The maps have dots moving around on them.]
[Dr Evangelos Pournaras says:] And here you can see how the system works. The agents, the appliances, interact in a distributed way to learn the schedules that optimize the power grid. The schedules that can prevent the power of blackout.
It looks like this network here.
[Dr Evangelos Pournaras shows a 3D metal model featuring nodes (round, metal balls) connected together with blue rods. It has a pyramid shape, with the network spreading out from the node at the top.]
[Dr Evangelos Pournaras says:] So we have the nodes which resemble the appliances at home and the purpose of this network is to coordinate the choices for the power demand.
This coordination happens in a bottom-up way starting from the leaves up to the root.
[Dr Evangelos Pournaras points to a node at the bottom of the model, then traces the connecting rods up to the node at the top of the model.]
[Dr Evangelos Pournaras says:] And then the root initiates a back propagation process that reaches again back to the leaves.
[Dr Evangelos Pournaras points to the node at the top of the model and traces the connecting rods to the node at the bottom.]
[Dr Evangelos Pournaras says:] The process repeats, and the power demand is regulated and the blackouts are prevented.
Via collective learning several highly complex computational problems combinatorial in nature can be solved efficiently while aligning with citizens social values.
[The screen shows someone using their mobile phone to adjust their home lighting, then a roof covered in solar panels.]
[A voiceover says:] By crowdsourcing several management and regulatory operations back to consumers and producers, system operators and utilities can significantly reduce their costs.
Collective learning promises a new decarbonization and sustainability pathway with active citizens’ engagement.
[The video ends with a screen showing the EPOS logo, and links to the EPOS website, epos-net.org and email address, mail@epos-net.org.]