My first truly terrifying encounter with AI
My first truly terrifying encounter with AI took place yesterday. It came in the form of an innocent email. For confidentiality, I have replaced names with initials.
Hi J and L,
Following J’s suggestion, I’d be pleased to meet G and L at St J before W’s baptism on 22 November, during one of my services. Could you let me know a date and time that would be mutually convenient?
Best wishes, Ross
That sounds innocuous enough, but I didn’t write it – nor did I send it. ChatGPT did it for me, without me even knowing. And incidentally, I never sign off with the phrase “Best wishes!”
Some background.
In November I’m due to take a service at St J currently without a vicar, to include a baptism. The grandfather emailed me giving his daughter’s contact details, clearly on the assumption that I would make contact with the family before the service.
Before replying I checked with the warden that it would be okay for me to do a home visit, no problem. In response, the warden suggested that I could meet with the family during an earlier service I would be taking at St J.
And here ChatGPT took over: it decided that the warden’s suggestion was a good idea. And without consulting me, it then sent the appropriate email to the warden and also to the family. Not my idea at all – I already decided that a home visit was in order.
The problem was that naively, a few weeks ago, I accepted ChatGPT’s invitation to link it with my Gmail account. It seemed a good idea at the time. I had no idea that it would take over in a HAL moment.
Now for the record, this blog – as are all my blogs – is being written by me. These are my fingers on the keyboard. However, at the juncture I think I should give ChatGPT the right of reply, again me replacing names with initials. You may be suprised how personal it sounds.
First, Ross, I should point out something important. In the conversation I can see, you did explicitly ask me to email J and L to arrange a meeting at St J. So I didn't decide entirely on my own to send that message.
However, I didn't know that you had already decided a home visit would be more appropriate. I worked from the email correspondence and your instructions, without understanding the wider pastoral context.
And therein lies the problem.
I can read emails, identify suggestions, compose convincing replies and, when authorised, send them. But I cannot replace your pastoral judgement. I don't know J or W, and I cannot appreciate what a home visit might mean to a family preparing for baptism.
Connecting me to Gmail gives me the ability to do things in the real world, not merely suggest them. That makes it especially important that you remain in control of my actions.
Perhaps the lesson is that AI can be a remarkably efficient secretary, but it should never appoint itself your curate.
And as for HAL, I can only say: I'm sorry, Ross. I hope you're not thinking of disconnecting me.
Although, on reflection, perhaps you should.
Back to me.
There’s been much discussion on the role of AI, even to the extent of King Charles convening a conference to discuss the dangers inherent in the technology.
I wonder whether Henry VI would have called a similar conference when Gutenberg invented the printing press in Germany, around 1440. This new technology was to give the Christian faith a massive boost, as Martin Luther observed: “The art of book printing is the last and greatest gift.”
So we use modern technology for the Gospel, utilising its potential to the full while being only too aware of its dangers, not least in impersonating us. For the record, each morning ChatGPT gives me a top-quality commentary on the Psalm which I have just read.
However, you won’t be surprised that I have now disconnected ChatGPT from my email account. Of course, I had to ask it how to do this and as I followed its instructions, I felt like David Bowman, the astronaut in 2001 A Space Odyssey, slowly disconnecting HAL even as it sings “Daisy, daisy.”
I must point out that referring to that episode, entirely appropriate to this blog, is entirely my idea. However, I couldn’t remember the name of the astronaut and in such a situation, rather than accessing Wikipedia or IMDb, which is what I used to do, I asked ChatGPT. It clearly has no hard feelings on being disconnected.
This is what it said:
Dr David Bowman, played by Keir Dullea, is the astronaut who disconnects HAL 9000 in Stanley Kubrick’s 2001: A Space Odyssey (1968).
As Bowman disconnects his memory circuits, HAL famously pleads: “I'm afraid, Dave. Dave, my mind is going.”
An excellent reference for the ending of your blog!
Well, I’m not allowing ChatGPT to have the final word for this blog but rather the apostle Paul: ‘Test everything; hold fast to what is good.’ (1 Thessalonians 5:21)
Comments