S2 EP26 Can countries strategically plan for the future? with Peter Nelson

My company developed Disclose AI, which is a chatbot where we set up, we curated the library of information that we share with the people, and we trained a model on Creo, Menda, and Temna, so the local languages.

We introduced speech-to-text, text-to-speech features, so they can ask any sort of questions about the process only, that is the Strategic Environmental Assessment, SEA 2050, and also provide feedback through that channel.

We also included an email address for general feedback communication, and the social media team has used several AI tools to generate images or edit images that we use to share information about the process.

Obviously, the first stage was the scoping stage, where we had a draft scoping report.

There it is.

That was shared for consultation.

Various snippets and findings of these reports were incorporated into the social media content strategy.

And in addition, we were scraping comments on Yes, we were collecting comments from our social media platforms from people to do a sentiment analysis.

We used AI machine learning algorithms because it was a large number of feedback.

In addition, we also had the WhatsApp channel.

where people often left voice notes.

So that was one of the reasons why we decided to train the model on the local language, because it did not recognize, you know, that was in 2024, there were efforts to sort of like include the South in like AI.

So there are pre-trained models on Creo, but those are more the text-to-text, not text-to-speech or speech-to-text.

So we gathered material, we used some of our local team to send us voice notes so we could fine-tune this product, and that helped us transcribe and translate messages from the people.

But the most interesting part for me about the AI stuff is this agentic AI government planning tool that basically absorbed all these data from publicly available sources, um you know national statistical office covering many many sectors from like mining agriculture healthcare tourism environmental degradation flood risk water management energy you know it took us a lot of brain power to to think about everything so that we don't miss out we set up these data sets and then we used open source deep learning and machine learning algorithms.

We fine-tuned some, we had to build some, because as you were saying, these topics are complex.

I love the way you've managed to explain that with only one buzz, because when Ildiko first started talking to me, I really struggled with some of the definitions and the process that's involved.

But I want to give a practical example, because this is the way that I try to explain it when we're actually face to face in the communities.

And that is the fact that Sierra Leone currently has 9 million population.

And the reason we talked about 2050 is because it will grow to 15 million.

It's a 60% increase.

It's another 6 million mouths to feed, another 6 million children who will have to enter schooling in that time.

and the education standards, virtually all of the parameters that in the Western world we would just accept as norms don't exist.

So if you talk about how many schools are going to be needed for six million children.

And where.