The Hitchhiker’s Guide to AI: The African Edge by Arthur Goldstuck
News24’s Book of the Month for August, The Hitchhiker’s Guide to AI: The African Edge, reframes the conversation about AI through the lens of human roles and merges global advances with distinctly African realities. From farmers using AI to track bee movements, to school pupils guided by WhatsApp tutors, to musicians experimenting with machine-made beats, this book explores how everyday people across the African continent are shaping – and being shaped by – the rise of machine intelligence. Rather than obsessing over distant futures or Silicon Valley breakthroughs, this book brings AI down to earth. Through the eyes of local and international teachers, coders, executives and artists, it tells the story of Africa’s AI moment – not as a catch-up game, but as reframing the global narrative.
Arthur Goldstuck is the author of 21 books and the founder of World Wide Worx, Africa’s leading technology market research organisation. This is his second book on AI.
In this excerpt, he writes about how AI learns to interact with humans.
BOOK: The Hitchhiker’s Guide to AI: The African Edge by Arthur Goldstuck (Pan Macmillan South Africa)
Natural language processing
Chatted with any interesting computers lately? Your knee-jerk reaction is probably that you don’t talk to machines, but you are also probably deluding yourself.
For example, most people nowadays find that the easiest way to get account statements, if the option if offered by your institution, is to call up a menu on WhatsApp and follow a sequence of instructions. Sometimes, instead of a number from a menu, you are asked to type in what you want. Miraculously, maybe, the machine gives you what you want. Many companies, ranging from Discovery Health to Mercedes-Benz to MTN to Vodacom, are using more advanced chatbots to do the same thing.
The results tend to be disappointing, as most of the chatbots are designed merely to pull answers from an existing menu on a standard website. But they are getting better all the time.
More and more of us are succumbing to the lure of voice assistants on our phones and smart speakers, asking Siri (Apple), Google Assistant or Gemini (Android), Celia (Huawei) or Alexa (Amazon) for directions or to play a song. Again, the machine obeys. Unless you are in a car and talking to the clunky built-in voice command system, in which case the machine almost always gets it wrong. But connect Google Android Auto or Apple Car Play to the car, with their access to the latest voice technology, and it is suddenly obeying your spoken commands.
How?
Thanks to natural language processing (NLP), a system for programming computers to process and generate text, speech and other forms of human language. The idea is to teach computers to understand and communicate with humans using the language we speak or write. In short, it’s all about bridging the gap between human language and computer language.
But how?
It starts with data. Vast amounts of it. Computers are fed with text from numerous sources and programmed to analyse and study the text to learn patterns, grammar and meanings behind words and sentences. Once it has passed English – or any other language – in this school, it is ready for any form of language processing.
NLP can automatically summarise a long article into a short paragraph, translate sentences into other languages and analyse the sentiment in social media posts. ChatGPT told me that sentiment analysis is ‘like the computer reading your mind’, but that’s just wishful thinking. It merely matches a large dictionary of attitude-related words to a set of rules that indicates whether a word is positive, negative or neutral.
When it works, it helps businesses gather insights about how people feel about their products or services. But think of a young person describing something as ‘sick’ or ‘wicked’. If the system is not programmed to pick up on slang, it will interpret a great positive as a negative. Once again, it is a case of: human, beware.
NLP algorithms use various forms of artificial intelligence, including machine learning, to break down sentences, look for keywords and analyse the context to generate meaningful responses. As the algorithms improve, and computing power improves, NLP improves.
Before long, the science-fiction-like promise of instant and automatic translation of languages for travellers will be an everyday reality. With luck, but don’t hold your breath, large companies’ customer services will be transformed.
Intelligent automation
Intelligent automation combines the power of automation and AI, with a promise of bringing efficiency, accuracy and intelligent decision-making to business processes. Left to itself, this form of AI has the power to wreak havoc, so let’s first look at the nightmare before we address the dream.
On 6 May 2010, long before AI had entered common use in business, Wall Street was rocked by an event known as the ‘flash crash’. A combination of market conditions and high-frequency trading algorithms, designed to execute trades at extremely fast speeds, resulted in a flood of automated sell orders that led to a cascading effect and a sharp fall in prices.
Because there was an insufficient number of buyers to counter the selling, some stocks briefly traded at absurdly low prices, as low as a cent or a fraction of a cent. Aside from widespread panic, it result ed in the temporary loss of billions of dollars – before the market rebounded just minutes later.
ChatGPT tells us: ‘The incident underscored the importance of monitoring and managing the risks associated with automated trad ing algorithms. It highlighted the need for robust risk management mechanisms, circuit breakers, and coordination between market participants and regulators to ensure market stability in the face of rapid algorithmic trading.’
Now that you’ve been warned, you probably won’t rush into intelligent automation, but let’s pause and take a breath.
At its core, intelligent automation automates repetitive and rule based tasks that were traditionally done by humans. This means it can streamline routine operations, save time, and reduce errors in data entry, invoice processing or report generation, for example.
It is one of the categories of greatest impact of AI on business, not only because of what it can do, but what it allows business decision makers to do: free up their time for more strategic and value-added activities.
The key is in that word that is at the heart of this book: ‘intelligence’. With AI added on, machines go beyond simple automation to analyse vast amounts of data, recognise patterns and, ultimately, make decisions.
Intelligent automation plays a crucial role in fraud detection, risk assessment and financial analysis, making it a powerful tool for financial services organisations. It is at the heart of a new generation of insurance companies, like South Africa’s Naked, which uses AI at every stage of the customer journey, from getting a quote and signing on to making a claim.
But the possibilities go far beyond moving money. The ability to process large volumes of data quickly, and identify anomalies, predict trends and make informed recommendations, offers efficiency and minimises risk to any business that produces a large amount of data.
By leveraging AI algorithms, machines can analyse data from multiple sources and provide insights for strategic planning, market analysis and forecasting. The integration of automation and intelligence in business processes can also allow collaboration between machines and humans.
This means it enables businesses to improve productivity and allocate human talent to more creative and critical tasks that still require people’s touch and expertise.
Behind the scenes, intelligent automation feasts at a smorgasbord of AI tools, including machine learning, natural language processing and computer vision. That makes for better results over time.
As these improvements multiply, they can have the very opposite effect of the likes of a flash crash. They promise a boost in value that is sustainable, measurable and, for the humans who keep their jobs, deeply satisfying.
