RefGPT: the solution?

Yesterday an article about RefGPT appeared in the RD, an alternative I had already come across earlier. Because the makers and I share the same concern, we got in touch with each other. In addition, I've been asked several times what I think of the initiative. Here's a short recap.

First of all, I find it praiseworthy that there is action: a plan, an idea. And if there's one thing I don't want, it's polarisation bringing that train to a halt again. At the same time, I do see some caveats that, in my view, were left underexposed in yesterday's article due to a lack of journalistic probing.

1. It is stated that RefGPT "barely hallucinates". This cannot technically be substantiated, because the nature of every AI is that it works with probabilities. Even with RAG systems, Azure as cloud platform, source referencing and filters, hallucinating remains possible.
2. The claim that RefGPT is 100 percent reformed-accountable is too absolute, theologically and substantively. Summarising and selecting is always interpretation, and there is no single "the reformed view". You can, for example, think differently about a text like Galatians 6:4.
3. The statement that RefGPT is a full-fledged alternative to ChatGPT also isn't quite right. RefGPT is deliberately limited and functions as a shell around an existing language model. It cannot, for example, generate images.

More caveats could be raised, for example around copyright of teaching methods. My main objection, however, is that AI has by now become deeply embedded, certainly among secondary school pupils. About 90% of those pupils already use AI (source: scholieren.com), and so far there's no reason to assume this is different among reformed young people. The question then is whether an alternative like RefGPT will take up enough space. Isn't investing in critical, Bible-aware AI literacy more thorough and fitting, also in light of the AI Act?

Finally: as far as I'm concerned there are certainly possibilities for RefGPT, but more limited than presented in the article. I see opportunities, for example, as a practice tool in primary education, and then exclusively within the school context.