Just How Far Can AI Go? We Asked an Expert to Find Out

While the threat of increased economic disparity is a great concern to Professor Marwala, he’s quick to point out how AI could be used for good, citing evidence of how it could be used to prevent buildings from collapsing.

He talks of how his grandmother used to make clay pots: “She would tap each pot and listen to the sound and this told her whether it was strong or weak. Using sound to test quality, is what engineers call non-destructive testing and it’s applied in aerospace engineering to assess whether planes have body cracks.

“The sound my grandmother was listening to is what we call vibration data in engineering and this can be processed by artificially intelligent machines.

“At the University of Johannesburg, we offer courses in vibration analysis, signals and systems, thermodynamics and artificial intelligence. These are necessary to take the framework I was taught by my grandmother and use it to monitor the safety of buildings and bridges.

“Data acquisition devices are embedded on buildings and bridges and data is relayed to an artificially intelligent machine, which analyzes it and decides if there is any danger of a collapse. In case of imminent danger, automated messaging can be used.”

Interestingly, one of the obstacles Professor Marwala envisages for the growth and development of AI around the world is the fact it will have to adapt to different countries and cultures. To give an example, he mentions how his smart speaker at home insists on saying his name in an American accent.

He says: ““The reason why this happens is because the data that is used to train this device is largely collected in North America and not in South Africa. For us to make this device not to be biased we will have to record all words and associate them to meanings in all of South Africa’s 11 official languages and incorporate these into this Google device.”

These developments will happen eventually - Google Translate has begun to feature African languages, including Swahili and Zulu - but it will take time. Some languages and dialects will also be easier than others. Professor Marwala cites the example of isiXhosa, a Bantu language which AI has found difficult to learn.

As isiXhosa consists of a mixture of clicks and Bantu language, AI interprets the clicks as background noise and eliminates them when attempting to understand or translate what has been said.

Professor Marwala says: “Nevertheless, isiXhosa is an important language locked out of the fourth industrial revolution. We ought to develop a new method that will not treat these clicks as noise but as an integral part of the language.”

Language isn’t the only area where Western-developed AI systems fail to work effectively for consumers elsewhere in the world. For example, at the University of Johannesburg, Gugulethu Mabuza-Hocquet completed her doctorate on designing algorithms capable of understanding that people of African descent have a sharper difference between the pupil and iris of their eye.

Professor Marwala says: “These algorithms, therefore, allow biometric systems based on the iris of the eye not to implicitly discriminate Africans in favour of Europeans.

“The next step should be to develop better algorithms that understand the Xhosa language. Taking our languages into the digital and the fourth industrial ages is our responsibility. We cannot just import technology, such as speech recognition machines, but we should adjust them to our particular environments.”

Are you fascinated by the possibilities of AI and wondering how you could follow in Gugulethu’s footsteps to tackle some of the most pressing issues facing this fourth industrial revolution? Click here to find out more about studying at the University of Johannesburg.