The research focuses on how the “new” emerges in social and technological systems and how humans and machines explore the space of possibilities and find new solutions.
The "Creativity, Innovation and Artificial Intelligence" team is the newest team of the lab. It started its activities at the end of 2017 and it focuses on the investigation of the processes behind innovation and human creativity and their interplay with the most recent advances in Artificial Intelligence, Machine Learning and Inference methods. From this perspective, the team is playing the role of a gluing interface for all the other activities of the lab. This is inline with a general trend: that of breaching the walls between teams and creating opportunities for fruitful exchanges and cross-fertilization. In the very spirit of creativity, this will help in seeding new ideas and let them blossom in a fertile breeding ground.
Towards a science of the new
Historically the notion of ‘the new’ has always posed challenges to humankind. What is new often defies the natural tendency of humans to predict and control future events. Despite very original mathematical constructions proposed so far, we think a coherent theoretical framework to grasp the notion of space of the possible is still lacking. Given this context, the research aims at providing a coherent and self-consistent mathematical formulation of the space of the possibilities - which includes its structure and its restructuring while it gets explored - and a mathematical modelling of the way systems - biological, technological, social - explore it at the individual and collective levels. Mathematically, the team faces extremely challenging problems connected to the following facts:
- we do not know the topology of the space of possibilities, nor how to extract it from the data;
- we do not know how this structure evolves over time at the individual and collective level;
- we still need a coherent and self-consistent mathematical formulation that, beyond explaining stylised facts (statistical laws, correlation and triggering effects, etc.), is able to cast concrete predictions to be grounded on actual data. Specific goals can be summarised as follows:
- establish an operational definition of space of possibilities and devise methods and tools to chart it;
- develop a self-consistent theory of the emergence of the new as guided by the different active principles;
- identify the best thriving environments for creativity and innovation with concrete applications in areas like education, research and business.
The application side of this research also includes the development of new recommendation strategies for the exploration of the space of possibilities (e.g., which products areas will likely to drive the market), the evaluation of new avenues and the quantification of innovative behaviours and the prediction of future success of novel elements.[click to read more]
Unveiling the strategies of creativity
This research line is focusing on unveiling of the strategies of exploration of the adjacent possible in many different systems (social, biological, technological). The goal is pursued by adopting and developing data science techniques on datasets mirroring the emergence of novelties in very different kind of systems, including dedicated experiments involving human creativity. From all these datasets, the spaces of the adjacent possible for the studied systems will be reconstructed and analysed, especially from a topological point of view, both in their static features and in their dynamic evolution. Crucial attention is devoted to the role of creativity in the exploration of the adjacent possible, in order to single out the most effective strategies of exploration. The data science approach described so far will be paralleled by the realization of actual experiments involving people, both through online gaming and open events (see for instance: www.kreyon.net/kreyonDays) where it is possible to engage people in activities that challenge them to explore their adjacent possible and come up with new ideas, recombining existing ideas, effectively triggering some evolutionary dynamics of novelties (videos of several activities are available here: https://goo.gl/Ejdy14). All these elements represent the basic toolbox to deepen our understanding of how creativity works and identify the best heuristics to boost human creativity and foster innovation in the most diverse environments: science, art, technology, industry, etc.
Towards a creativity of machines
One of the most important themes in the next few years will be the understanding of the success of the deep-learning-like tools and the positive synergies theoretical schemes and data can jointly trigger. In particular, the efficiency of different algorithms in supervised or unsupervised learning has been argued to respond to some properties of the natural, inferred data (as natural images and natural language), properties among which symmetry and a certain hierarchical structure, so that deep (or multiple-layer) learning algorithms can be interpreted as a form of the real-space renormalization group techniques. Even more important will be thinking about the way in which inference tools can be enriched to take into account processes where innovation is present. When brand new events are present, inference schemes have to be revised and theoretical modelling becomes crucial (See the paradigmatic example Volvo_vs_kangaroos: https://goo.gl/PJhKcs). From this perspective, the goal of this research is twofold:
- opening the box of deep-learning algorithms to understand what has been defined as the “Unreasonable effectiveness of learning neural networks” (see https://goo.gl/qkQt3Y). Methods coming from statistical physics of complex system could be very useful to this end.
- exploiting the knowledge of the way the space of possibilities is explored to conceive the next generation of artificial intelligent algorithms able not only to treat novelties but also to predict them, incorporating creative behaviours in a dynamical way and helping us to drive our technological and cultural exploration of the adjacent possible.
This will be the starting point for a deep understanding of the very notion of creativity for machines, pushing their evolution to the next step (this is very much in line with https://goo.gl/7KXjwC). The applications are very many, encompassing the whole spectrum of arts, sciences and technologies.