James Browning
Text-based AI models, formally known as Large Language Models (LLMs), have been enjoying widespread media coverage this year.
By being trained on arguably the largest set of text data assembled (made possible in large part by the internet), the models do an impressive job of mimicking humans on a wide range of tasks.
They can write essays, hold back-and-forth conversation, and brainstorm. Importantly, recent language models have also proved to be effective at writing and explaining computer code.
The potential applications of the technology are too wide-ranging to detail here, but a useful case that shows great promise is AI tutors.
General-purpose tools like ChatGPT are good at helping users understand topics but this scratches only the surface of what is possible.
Khan Academy, a non-profit whose high quality free tutorials have been used by millions of students worldwide over the past 15 years, recently demonstrated its in-development tutor chatbot.
Rather than simply giving students the answers to their problems, the chatbot can converse with a student and guide them to an understanding of how to solve the problem themselves.
While AI tutors could be effective in many areas of learning, from writing skills and literature to maths and physics, the most valuable opportunity for South Africa lies in their ability to teach code.
AI tutors will not be replacing teachers in basic education any time soon (and there are good arguments for why they never should), but they may prove to be useful for technical disciplines like coding and broader IT work.
India makes for a good example of how digital technology skills can provide significant employment in a developing country. This may be even more relevant in a South African context, where failing infrastructure hamstrings attempts to create economic growth in manufacturing and material-based sectors.
Digital work has substantially lower resource costs. Basic computers can be effective coding tools and, unlike industrial machinery, can run during load shedding, using relatively inexpensive back-up power.
Crucially, digital work is inherently global. A web-based start-up in Durban can reach customers worldwide without the cost of exporting physical goods.
This allows businesses to access customers in wealthier countries, who often have more disposable income and stronger currencies. Five dollars goes further in South Africa than in the US, and trading internationally can efficiently funnel wealth from the richer Global North into the local economy.
AI can be an invaluable tool for giving South Africans coding and technology-related skills, which not only make them employable but, most importantly, empower them to become entrepreneurs.
Robot tutors are always available, have infinite patience and can access an unbeatable breadth of knowledge. This means they can start teaching subjects from the ground up, which is essential for a population with a low level of computer literacy and which has often been failed by the basic education system.
Unlike a textbook, an AI tutor can adapt to the pupil and help them at their level of knowledge, even if that means going back to foundational maths and problem-solving skills.
To be clear, training by actual humans should never be undervalued, and skilled teachers would be essential in the process of developing the AI tools.
However, AI tutors can provide personalised education at a huge scale and a fraction of the cost. And training South Africans in digital skills specifically has the potential to be a key component in addressing unemployment and a stagnating economy.
James Browning is a freelance tech writer and local music journalist.
BUSINESS REPORT