Creations tend to inherit its creators' traits.

Guest lecture:

AI for social impact

The area of "AI for Social Impact" or "AI for Social Good" is fast emerging as a new field of scientific and technological endeavor in the use of AI to solve wicked societal problems.

While there is a growing movement to direct AI usage towards 'good' uses (i.e. AI4Good), towards the sustainable development goals, for example, it is time to move beyond that and to address the sustainability of developing and using AI systems in and of themselves.

Unfortunately, the benefits of AI do not come without risks. According to the World Economic Forum (WEF), left unguided, AI "has the capability to accelerate the environment’s degradation."

One way it could do this is through its energy consumption because many AI systems require enormous amounts of energy to perform. Some reports say global emissions from cloud computing emit more carbon than commercial airlines and a recent study showed that, by 2027, the AI industry could be using as much energy as a country the size of the Netherlands.

Case studies and achievements

Case study: GraphCast

Describe

(image)

Graphs and graph representation

To quickly recap, graphs are a mathematical structure used to represent relationships between entities. It consists of nodes/vertices that represent entities and edges that represent relationships or connections between them. Graphs are generally sparse structures where only a few subset of edges are present.

Examples

Types of graphs

Graph representations

• Adjacency matrix

Permutation of node indices

 Graph neural networks

A graph neural network (GNN) is an optimizable transformation on all attributes of the graph (nodes and edges) that preserves graph symmetries (permutation invariances). Used in predicting nodes, edges, and graph-based tasks, GNNs are highly influenced by CNNs and graph embedding. Similarly, GNNs are applied to graph structure (grid of pixels) to predict a class.

Tasks

Permutation invariance and equivalence

Permutation equivalence

(image)

Text

(continued image)

GNNs and permutation invariance/equivalence

Graph convolutional networks

Graph convolutional networks (GCNs) are

(mixed image)