The aroma of filter coffee welcomed delegates to April’s “data@breakfast” seminar at the Senate Chamber, Westville campus. Speaker, Dr Maria Schuld introduced her diverse audience to the behind-the-scenes world of machine learning.
Schuld, who holds Master’s degrees in Political Science (Free University, Berlin) and Physics (Technical University of Berlin), as well as a PhD from UKZN, is a postdoctoral researcher at UKZN’s Centre for Quantum Technology.
She defined machine learning as the intersection of statistics (generating and analysing data), computer science (developing algorithms and hardware) and artificial intelligence (giving machines human-like abilities) – all underpinned by a foundation of mathematics.
‘Machine learning involves making machines perform tasks without being explicitly programmed,’ explained Schuld. ‘It is data-driven decision making, recognising and reproducing patterns in data, and analysing “human-readable” data to produce “human-like” outcomes.
As citizens, we have to know about machine learning systems because they will define what happens in the future and where the money goes,’ she added.
The presentation conveyed three messages: machine learning is getting better; machine learning isn’t rocket science; and machine learning reflects and affects our society.
‘In the 1950s, machine learning was human fed,’ explained Schuld. ‘By 2006, only the algorithm was devised by the human practitioner; everything else was done by the machine.’ She pointed out that machine learning is used constantly in daily life, for example, text prediction on cell phones and Google Translate on the internet. Both these services improved dramatically when machine learning advances enabled the data and correlations in the data to speak for themselves.
Schuld reassured the audience that machine learning is a “learning by doing” programme. ‘If you have basic programming skills, this world is open to you,’ she said. ‘Machine learning is quite a democratic field if you are given some starting resources.’ She explained that a number of resources are available on the internet such as fast.ai and kaggle.
One aspect of machine learning that particularly interests Schuld is the ethical debate surrounding artificial intelligence (AI). ‘These days, ethical debates in AI are a lot more immediate,’ she said. ‘Machine learning reflects and affects us as society. How do we move machines away from the bias that is inherent in the data? How do we ensure that machines avoid data pre-fed prejudices?’
Schuld said that AI and machine learning is particularly interesting in the African context. ‘More diverse people tend to be involved and the topics under discussion are not purely commercial, but are far more influenced by broader societal issues. This is what makes machine learning in Africa so relevant and exciting.’
UKZN’s monthly “data@breakfast” lecture series are the brainchild of Pro Vice-Chancellor for Big Data and Informatics, Professor Francesco Petruccione, and are a platform to discuss ‘topics at the forefront of big data, artificial intelligence, quantum computing, machine learning and more.’
Big Data and Informatics, which offer computing solutions for the future, is one of the four Research Flagships identified as critical to UKZN’s current Strategic Plan. The others are Social Cohesion (addressing inequality and promoting nation building), African Health (saving lives), and African Cities of the Future (most liveable cities).
Words and photograph: Sally Frost
