Journalist, writer and curator of the Wall of Great Africans
What you need to know:
ChatGPT, and other AI-powered chatbots, can write music in a chosen style.
Beyond the dazzling results, these chatbots' raw material is relatively basic information that is on the internet.
Digitisation has been one of African media's biggest challenges.
In early November last year, I spent days on a writing project trying to figure out how many islands are in Lake Victoria and how many Kenyan and Tanzanian-owned islands were in the Indian Ocean.
There were so many links and, with varying figures brought up, I gave up and went with something wishy-washy.
Shortly after, news of the AI-powered chatbot ChatGPT was all the rage, so I gave it a go. In about a minute, I had pithy answers to all three questions, with no messy links cluttering the result. I did it again. There are 985 islands in Lake Victoria, came the result, a slight variation from the first result I had got. It said Uganda's Sese Islands make up 84 of these islands, which it didn't bring in November, indicating just how much ChatGPT is evolving.
Tanzania has more than 2,000 islands in the Indian Ocean, it said. Kenya is island-poor by contrast, with a miserable 30-plus islands in the Indian Ocean to its name.
ChatGPT, and other AI-powered chatbots, can write music in a chosen style, write code, do speeches, serve up essays, and do all sorts of other clever things.
Doomsayers aroused
The doomsayers have been aroused, declaring that so many jobs in journalism, coding and law, to name a few, will soon be killed by the chatbots. Perhaps not. In my case, if I had started my quest about the islands on the chatbot, I would have saved myself days of frustrating labour.
Those are days I could have used to learn how to weave a basket, finish reading a book, or find a more imaginative way to write the article. From an East African — or better, African — perspective, the big issue with the coming AI-powered chatbot age is not the jobs that will be lost but the ones that will not be created.
Beyond the dazzling results, these chatbots' raw material is relatively basic information that is on the internet and in the recesses of the web that many search engines don't get to. It is just that they trawl the internet and process the information in a much faster and more efficient way.
Not a rosy picture
Therefore, for Africa, it's still about the kind of foundational about it that it is on the web. It is not a rosy picture. I read a while back, for example, that there are more articles on Paris on Wikipedia than there are on all of Africa. To get the specifics, I queried ChatGPT. It came back with the answer: "According to Wikipedia's statistics, there are currently over 60 million articles on Wikipedia, with the majority of them being in English. Out of all these articles, only about 1.4 million are about Africa, while there are over 2.7 million articles about Paris. Therefore, there are more Wikipedia articles about Paris than about all of Africa."
If we are to get a premium from these AI chatbot technologies, that is what we have to fix. Although Africa comprises nearly 17 percent of the world's population, only 5-10 percent of the content online is related to the continent. In the next two to three years, considering that the rest of the world won't stand still, African content on the web needs to grow by more than 200 percent to provide enough fodder for AI. Otherwise, another inequality front will open there.
At the pace of development, for example, by 2025, a smart producer should be able to do a historical documentary on Africa with the help of only one or two other technical people. We need a massive digitisation push to bring online analog archives and a large number of photographs, videos and text on Africa's shelves and drawers that are still in ancient format.
Real story
A while back, I heard that one of the most ambitious digitisation projects in Africa was being undertaken by the Rwanda Broadcasting Agency, among whose properties is Rwanda TV in Kigali.
On a visit to Kigali last year, I went to RBA to see with my own eyes what was the real story. I would never have been prepared enough for what I found.
It was a massive operation run by a team of Indian experts and Rwandan techies. They were nearing the end of the digitisation of nearly 60 years of audio, images and everything that ever came through their TV and radio. Particularly remarkable was the granular annotation and optimisation for future searches.
Digitisation has been one of African media's biggest challenges. Kigali might be sitting on the most successful execution of it in Africa. RBA officials recently said there are now delegations trooping in from around the continent and the world to see how it is done, and they might even earn a buck for their expertise. These are the fellows who will come out winners at the other end of the AI chatbot tunnel.
Strategically, then, Africa must pour serious money into digitisation. And then it must tinker with education to produce students who ask more creative questions. The Good Book says, "You can't make a silk purse out of a sow's ear." If you ask AI how to make a hat, it will tell you how to make a hat. If you ask it how to fly to the moon, you will get the appropriate answer. There has never been a time when a people's future rested on the kind of questions they ask like today.
Charles Onyango-Obbo is a journalist, writer, and curator of the "Wall of Great Africans". Twitter@cobbo3
