Artificial apologetics?

In a study of 1,016 job categories, only 36 were shown to be largely unaffected by the rise of Artificial Intelligence (AI). These included ‘jobs’ like athlete and dancer. What about evangelists and apologists?

You can ask ChatGPT to “prepare a talk on Why God Allows Suffering” or “present a case for the reliability of the Bible” and, at the click of a button, get some pretty good suggestions. Artificial apologetics has arrived.

Extend it a little further, or add a few further prompts, and you get impressive results: ‘Prepare a set of 12 Bible Study questions on why God allows suffering from the NIV in line with an evangelical statement of faith suitable for a group of non-Christians who are interested in church but are not yet familiar with the Bible.” If you have not tried this kind of thing with ChatGPT then do so, and you may be surprised.

Persuasive pitfalls

It feels like cheating. Of course, at one level it is no different than taking a published Bible Study book off our shelf and using that. But, at a deeper level, we are concerned that no human being was directly involved in creating the content.

What about mistakes? AI software is getting better by the day, but there are curious errors. Random, incorrect references can appear – called “hallucinations”. Misunderstandings of the context of a word can send AI down rabbit holes. We should also be concerned about other agendas that can skew the information. Deepseek, a Chinese equivalent to ChatGPT, looks quite innocent until you converse about matters that might be sensitive to the Chinese government. Sometimes it will flash up a useful answer but then it mysteriously disappears from the screen to be replaced by a discouragement to discuss such topics.

AI is also creating problems for copyright law – and the environment. It is estimated that 16 queries to ChatGPT uses the equivalent electricity of boiling a kettle.

Artificial evangelism

I asked ChatGPT whether it was more or less likely that God exists – it thought it less likely. The most it would admit was that a “deistic or unknown form of God could theoretically exist, but the specific gods of major religions – especially personal, interventionist ones – seem unlikely given the available evidence.” Then I pursued the conversation by giving reasons to believe in God. At one point in the flow of answers it paused, and a message came up: “Reasoned for ten seconds” (I will take that as a win), before it continued the exchange.

During a conversation like this you realise that AI is trying to find the answers that you want. Of course, ChatGPT has no beliefs of its own, so if you give an objection, it will generally reply with something like “That is a compelling argument” – AI is nothing if not polite. When an AI program was tested in the role of a medical doctor it scored highly, not just for the accuracy of medical diagnoses but for bedside manner. A study suggested AI can be ten times more empathetic than a human doctor, at least in verbal form. While that may give the illusion of being personal, it only really reflects the fact that AI is a tool that is trying to please the user.

Under the hood

ChatGPT is called a “Large Language Model”. What that means is that the system has been training on vast amounts of text data. It doesn’t store those billions of books and articles, but it learns the patterns in language and predicts what words we might like to read. It isn’t really thinking and has no opinion of its own. But it certainly feels like it has.

No one knows for sure where the development of AI will lead. The algorithms involved in this kind of processing mean even the software engineers are not entirely sure what AI is doing. That makes for the stuff of science fiction. There is a story told of an innocent office worker who instructs AI to maximise the production of paperclips. It has the sole purpose of finding the best ways of maximising the production of paperclips. Let loose with the task the entire earth ends up stripped of all resources, human life ceases, and all that remains is a vast ball of paper clips.

A servant not a master

So, what should our attitude be to AI today? It is a tool. In an excellent book by Ethan Mollick, Co-Intelligence: Living and Working with AI, there is a good argument that once we understand AI is a tool, it can be used to enhance what we do. Rather than asking ChatGPT to generate an argument for the existence of God, it is more useful to interact with it over the various ideas we have. Ask AI to assume the role of Richard Dawkins or Jordan Peterson and put your arguments to them. This can generate genuinely creative and interesting ideas and enhance our use of apologetics. Like so many inventions in the modern world, AI makes for a poor master but is a useful servant.

AI is here to stay. Too much reliance on it as a tool may well be damaging. Having access to information, as we do with the internet already, is not the same as understanding that information or being able to critically assess what we read. ChatGPT could add to that shallowness. We may get better at presenting ideas – a ChatGPT-written talk – but less able to understand those ideas. Alternatively, if we use these innovations as tools, they can be particularly helpful for exploring evidence, arguments and how we communicate our faith.