Dystopian and Utopian

How I am struggling with the ethics of AI

[A scared Dr Syntax and his Robot]

I'm in a state of gobsmackery if there is such a word. For 3 days I have been analyzing a series of published and controversial reports from all sorts of angles, the rigour of thinking, definitions, the tone of language, the rigour of research methodologies. It's the sort of thing that is very familiar to me, that I usually do by marking up documents, coding in some way and so on prior to writing it all up and drawing conclusions -- a kind of literature review/summative evaluation at the desk. But this time around, I have been using generative AI on the reports- all of which I had read or skimmed in the past. And in fact, some of the material, particularly the more advanced statistical stuff, I really don't have a personal handle on, but I do have enough experience to recognise what they are, but not be critical beyond a certain level.

I started working and questioning the quality and so on about one document. Based on the results, I asked for different forms of writeup -- as dot points, narrative, developing explanatory tables and so on.

Then I realized (how stupid of me), that I could work on multiple documents at once, and ask a very complex interdisciplinary and multilevelled question something that is impossible to a human being. Boom. Answers in seconds. And from that, I asked even more precise questions, engaged in a kind of debate and this is the shock, it made suggestions about what to ask next, or alternatives (all based statistically, on other's queries about data). Sometimes I could see errors in the answers, and I told it of the error and to correct, which it did. Or I added in more material or insights to consider. And I drew in some other academic concepts (which I was familiar with) and used that with the material.

The whole exercise has resulted in a powerful explanatory table and some precise explanatory sentences, as well as some pretty good summary documents from the incredibly long explanations i got. It is stuff that might have taken months to discover, and I might never have discovered it. I even got to play around with the style of presentation.

Now, I knew my data set to begin with, and I know a bit about research methodology, but this tool as taken me to a new depth of well, original knowledge -- all based on trillions of bits of data being interrogated for their relationship to each other and then put into line as words. It was easy for me to see when it drifted into bland generalizations because there was not data and it was trying to be helpful. I didn't get hallucinations.

What is original here, what is not? Do we just say, well, it's like running some formula for stats and getting the results and then running another test. Instead of the slide rule or using a calculator you have used the computer. Or is it really of a different quality here, it has done far more than come up with answers to my natural language algorithms? Are they my conclusions or the machines? Is what so offensive is that it more or less talks like us?

All this is outside the questions of putting researchers, writers, editors, out of work or the vast amount of energy it consumes. Or instead, will new forms of work come into being and it will go green? I don't have an answer, but I do know that in my work life the kind of work I did could not have existed before the invention of the internet, and within that period, communication changes also meant new ways of doing things. From duplicated letters stuffed in envelopes (sometimes taking days of work), to emails, to clunky voip, to zoom. Time and space disappear and not just here. It was all present in Bangladesh as well for me. Villagers were online. As much a we can be dystopian about social control and Mammon, we can also be utopian about the opportunities to manipulate it all for social good.

And of course do I paraphrase, and make changes as necessary to what turns up in an instant. This takes time, that valuable commodity. Alternatively, to remain at least textually original, do I use more or less what the box has produced and acknowledge textual non-originality. I suppose if you care about language, you will go for the rewrite, but if you go for the latter, does it matter, as long as you can show that you were the driver, so to speak?

I know that in the academic world there is a lot of thinking going on about this from an ethical point of view, but here, I'm just reflecting on my own experience. How are others coping?