Have you ever felt that Blade Runner’s scenes from 1982 are transforming into reality nowadays? Do you recognize the concepts of Star Trek: The Original Series from 1966 in the solutions or products of today? Or maybe even you yourself are developing one of them? Then – for sure – you are curious about Artificial Intelligence and where it will take our society. My interview with TEDxVienna speaker Ramin Hasani, Researcher on Machine Learning at the Vienna University of Technology, is aimed to present to us the status quo of the Artificial Intelligence world.
Artificial Intelligence from the computer science point of view
AI seems to be a hype nowadays. At the same time, if you ask ten different people in this room “What is AI?”, you will probably get six different answers. The rest of the audience will struggle with the definition without saying anything. What is your definition of Artificial Intelligence?
RH: Any agent that can process data and generate outputs which increase its performance on a particular task can be called an intelligent system. These systems – intelligent systems – can be either experts’ systems that are usually a set of rules that are hand-coded through bodies of knowledge, and can reason about the environment. Or – Intelligent agents can be learning systems; they attempt to improve performance progressively on a specific task by adapting their parameters, given a set of training data. Learning systems generalize well to unseen data, like humans. Both types of systems – expert systems and learning systems – are called Artificial Intelligence agents.
Artificial Intelligence to perform decision-making by increasing systems complexities
AI will enhance autonomy in future industries. In factories, AI can be deployed for efficient data processing and decision-making, controlling robots, machine’s health monitoring, multi-agent tasks, etc. Usually, as long as we have access to data from a system-under-test in factories, we can build an AI system to understand the system, ensure its safety, and improve its performance.
In the financial sector, AI is significantly beneficial, especially when it comes to delineating investment strategies and predicting market behavior based on complex and multi-dimensional criteria influencing it. AI has long been a toolkit used effectively in financial technology. Since 1956, when the application of Bayesian statistics became popular, it’s in use. In the 80s Expert system as Financial Decision Support systems came into play and became popular. Nowadays, machine learning systems play a central role in financial ecosystems, from customer data processing, to managing investments, to assessing risks.
Another sector in which AI can play an import role is the education sector. Education is changing towards a life-long learning model. Therefore people are continuously improving their skills in order to be able to cope with the technological advancements, achieved mostly by AI systems. AI will be tremendously beneficial in the process of life-long learning, in the form of personal assistant agents to ease the continual learning process based on individuals’ specific needs.
Artificial Intelligence in prospective research
What is the current focus of the prospective research on AI?
RH: There are many topics that I can think of. For example, fairness in AI: How can we design an AI algorithm that makes fair decisions? Fairness studies the impact of the AI systems on the society and the domain in which they work, as well as understanding how the AI agent comes up with a solution. Ethics in AI is another domain which became popular by the increasing of the learning capacities of AI systems.
On a more technical side, the field of reinforcement learning, which is learning by experiencing and interacting with the environment to maximize a reward function, has led to great achievements. This is proliferating to make better AI systems.
The title of your TEDxVienna topic is “Simple artificial brains to cover complex tasks” – why did you choose this topic?
RH: In my research, I develop mathematical models of the brain of small species and explore computational principles of how they achieve so much with so few computational resources. Together with my colleagues at TU Wien, IST Austria, MIT, and IMP Vienna, we investigate how we can create AI systems that work similar to the small species’ Brains. We then expand the application of these brain-inspired learning systems to real-life control tasks in an attempt to enhance the transparency of the contemporary AI systems.
If I think about AI and my motivation to work on AI, I can say that I try to turn the black-box models into white-box in order to increase auditability, ergo designing safer AI agents, in safety-critical applications such as self-driving cars.
Ideas worth spreading
It’s your first TED talk as a speaker – how does it feel?
RH: It’s exhilarating – at the same time, I am not nervous. I am excited about the talk and getting feedback from the audience. I would like to take it as it comes and look forward to the people’s reaction to my research. In my opinion, the most effective way to conduct and evaluate research is to spread your idea and precisely process the resulting reflections – so ideas are indeed worth spreading! Nice motto (Laughs).
Photocredits: Natalia Sanderson