AI in Africa...

I’m in Africa because it’s important to be in a place that gives you perspective. All the usual stuff when you travel, arrive at hotel to find there is no room, wandering about Dakar finding a taxi to a new hotel, being stopped by the police! In any case, we’re in this amazing city and you have to go with the flow.

E-learning Africa, run by the irrepressible Rebecca Stromeyer and her team, have been doing this for many years, because they think it matters. It’s easy to go to these big conferences in Europe and the US but what you get is a cyclopic view of the world, where we assume that everyone works in offices or at home on Zoom. The interest in the topic at our workshop was intense and I'm in a session on AI for language learning. as well as a huge formal debate on the role fo AI in Africa.

 

At this conference in these places, you are faced with the realty the rest of the world. Having been to Uganda, Rwanda, Namibia, Ethiopia and this year, in Senegal. It’s a big world out there and the worst conferences I’ve attended have been in two of the more hideous places on our planet Orlando and Vegas, the first seems like America embalmed, the second some sort of theme park from Hell. Come to Africa and see real unvarnished world.

 

When this new Age of AI struck like lightening in November 2022, I had, for years been extoling the possibilities of its massive impact on learning and health. That impact is not about the already wealthy, the graduate generation in the Northern Hemisphere, it’s about the rest of the world. From the holy pronouncements on ethics and AI from Brussels, you’d think they’d have a little more humility, seeing as well under 10% of the world’s population. The US is only 4.25%.  They both need to get out more.

 

When you have 1 teacher and classes of 50, or 1 doctor for 10,000 people, you have a different perspective on life. Technology matters more here, which is why Africa leapfrogged the rest of the world on mobile – it mattered when your precarious employment depends on that next text message or that your fragile finances need to be executed efficiently and cheaply from the rural world in which you live.

 

I do believe that AI technology is finally the technology that franchises almost everyone. It has the ability to raise productivity to pay for the education and healthcare globally. It can deliver the sort of Universal Teacher and Universal Doctor technology that is finally on the horizon. A teachers that knows more than any teacher, in any subject, has the pedagogic knowledge built into its methods, that is tirelessly friendly and supportive, 24/7, anywhere in the world.

 

But what really matters is delivery in all languages. We live with the legacy of colonial languages. In Africa they are English and French, in South America Spanish and Portuguese. Even in the rest of the word, English has become the lingua franca. Imagine a technology that can teach and deliver healthcare in any first language. Here's the good news. That too is on the horizon. Not only is a wide range of language available from the get go, that number has and is expanding into the thousands. Far from ignoring minority languages it may save them. Generative AI is a modern Babel.

 

Generative AI through ChatGPT was launched in 95 languages. allowing one to, write, summarise, error check and translate between these languages. That is astonishing but that number has grown to include many more. This has been one of the most used functions with significant increases in productivity. It was also launched knowing 12 computer languages, allowing translation between them. One amazing feature of Generative AI that slipped under the radar was its astounding capabilities in speech to text and text to speech. Whisper from OpenAI was world beating and free. What a start.

 

On the day I flew to Senegal, Yann Lecun announced a new Multilingual Language Model that works in 4000 languages! There are only 7000 languages in the world, so that’s more than a great start, it’s into majority territory. Early days but the promise is now here, that AI will open up education, health and other areas of human endeavour to everyone, no matter where they live or what language they speak. Once we can all understand each other in real time translation, perhaps we have a chance of better understanding and co-existing with each other.

 

This really matters, as technology tends to be skewed towards English in particular but also the usual suspects of largely northern hemisphere languages. If we are to see AI as a genuinely global and liberating technology languages do matter. What these large language models have shown us is that language really does matter in learning. Wittgenstein and Vygotsky were right, it is fundamental to learning and intelligence, fundamental to being human. It is not that intelligence produces language but that language produces intelligence.

 

So let’s not allow the over-developed world get on their moral high-horses and hold back the technology that promises so much to so many. A vast army of amateur ethicists (anyone can be one by simply saying they are one) seem determined to hold us back with so much verbiage on the subject that you could train a Large Language Model on that alone. Unfortunately it would an aloof, finger-wagging, moralising bore.

 

In truth, for those most in need, who often find themselves furthest away from education and healthcare, its power and reach are potentially immense.