AI Interpretation (or – am I redundant yet?)

My inbox has been full of waffle about AI for a couple of years now. Apparently AI could be writing my podcast show notes, my blog posts and lots of courses and books for me to sell. All I need to do is spend 10 minutes entering prompts and it will make my fortune while I lie on a beach, moving only for the occasional basting with sunscreen.

Hm. (Or in other words – I have slightly less than no interest and none of that is ever going to happen.)

(I did ask ChatGPT to generate an image describing casting Yijing readings with AI to illustrate this post, though. I wouldn’t normally use AI-generated images either, but its output for this, pseudo-bagua and all, seems peculiarly apt.)

I’ve found it all quite easy to ignore so far – except, apparently, now AI interprets Yijing readings too. ‘Enterprising’ individuals are already offering AI-generated interpretations, and in some ways they’re a great deal better than I would have thought possible. AI, it turns out, can put together a human situation and ancient imagery quite intelligently, especially if you nudge it in the right direction.

This thread started by Remo Dentato, our much-valued expert on casting methods, gives some idea of what kind of nudging is required. He’s got his ‘YiShi’ bot to cast readings as well as interpreting them, and reliably using and quoting from Wilhelm’s translation. (He used the Wilhelm/Baynes translation at first – including in the example I link to below – but has switched it over now to a machine-generated translation from the German.)

Why is this worth doing, or even contemplating? Remo said he wanted to be able to tell the difference between AI and human interpretation, ready for the day when someone would start trying to sell the AI variety. Would we be able to tell the difference?

“The point is not if AI will get “human characteristics” (like our “soul” or “heart”), the point is how “good enough” it will be able to approximate those in our eyes. There is little consolation in saying: “yes, it sounds like a human, it acts like a human, but in the end it is not”. In the end if it will smell like a rose it will be considered a rose.”

True. Though this isn’t what worries me, so much. If AI can produce a genuinely good interpretation – actually help someone to connect with the oracle – then it doesn’t greatly matter to me whether someone mistakes it for a human being. It would be a bit alarming for me if an AI could do my job as well as I can, but it would be a net good for the world at large.

No – the worry for me is whether AI interpretations are actually any good. Are they going to help people, or are they going to mislead and confuse? Will they make possible more real encounters with the Yi, or fob people off with an empty imitation of the real thing?

What can AI do with readings?

I’ve been feeding some readings to AIs – mostly to Remo’s YiShi – to see what they come up with. The first thing I realise is that AI can work well with metaphor. It understands (or do I mean ‘understands’?) what imagery is, and it does a good job of finding applications for it.

A reading I recently recorded for the podcast, for instance (so I do have permission to share this, though the episode may not be published for a while), was about whether the querent, recently let go from her last job, should try to redirect her career path towards something more ambitious and fulfilling. She cast Hexagram 50, the Vessel, changing at lines 2 and 6 to 62, Small Exceeding – an interesting mix, I think you’ll agree!

I found I got the best results from ChatGPT by asking it specifically what a particular image symbolised. What about the Vessel? That, it said, could be the person asking, and “The Vessel ultimately represents the process of making your life and career a meaningful offering.” That’s really not bad, is it?

I did the same with the bird of Hexagram 62, and received a neatly-formatted list of meanings, with subheadings. Here’s the first of them, as a sample (I’d fed the reading to the AI in the querent’s own words, as if she were the one asking, hence the second person response):

“The Bird as a Symbol of Overreaching
The line, “The flying bird brings the message: ‘It is not good to strive upwards, it is good to stay down,'” serves as a warning. Birds soar high, but when they fly beyond their natural limits or recklessly into danger, disaster follows. In this context, the bird represents ambition or action that is too bold, too big, or ill-timed.
For you, this bird symbolizes the temptation to make dramatic changes in your career, perhaps driven by frustration or a desire for immediate resolution after being laid off. Hexagram 62 reminds you that this is not the time for grand flights—such as abruptly leaping into a completely unfamiliar path or taking on something too large to handle. Instead, the focus should be on staying grounded, paying attention to small steps, and moving carefully.”

Again, I think you’d agree, this is a reasonable interpretation. More subheadings followed:

• A bird in flight: messages and opportunity
• A bird between two realms
• The bird as precision and detail

Then came a summary, with a couple of suggested applications to the question. I was surprised and a bit disconcerted by how good all this was. I would have thought that deep understanding of metaphor was a uniquely human skill – but evidently not.

A full example reading

A. wrote to me to ask about AI readings and whether they are reliable (good question!), and has kindly given me permission to share his own example:

“My family is involved in cold chain logistics, mainly focusing on storage for chilled and frozen food products. Recently, there is a huge demand for cold storage space and we can hardly meet those demands. It is a good thing which means business is good but trying to meet all the demand is almost impossible and we can’t decline our customers’ demand. So I consulted Yi Jing asking, “What can I do today to increase my coldroom space short term to provide for my customers’ need?” and the result I got was hexagram 7 changing to hexagram 19. What does this mean?”

Not a question I’ve ever encountered before! Actually, it’s the kind of thing I can imagine someone simply asking AI, without involving Yi at all. But given this reading, what would you say?

‘The Army: with constancy.
Mature people, good fortune.
No mistake.’

‘The army sets out according to pitch-pipes,
Blocking strength, pitfall.’

‘Nearing.
Creating success from the source, constancy bears fruit.
Arrival at the eighth month means a pitfall.’

With Hexagram 7 in the reading, I would talk about getting organised, focussing on the problem, doing whatever’s necessary (all of which I’m sure A. has thought of already) – and about the importance of measure and being the adult in the room, not just throwing resources at the problem.

I generally understand the warning in Hexagram 19 about ‘arrival at the eighth month’ to be about concentrating too much on results, at the expense of the ongoing process of growth. So even though A.’s question was specifically about what he can do now, in the short term, I think this relating hexagram might be looking a bit beyond that. If this is not just a seasonal blip but a growth trend, he needs to ensure that the urgent short-term work doesn’t get in the way of long-term planning.

And the moving line… well, from what I’ve been able to learn about the military use of pitchpipes, they didn’t only communicate orders, they were also a means of discovering the true condition of each army and predicting outcomes. The orders need to be in harmony with the army, and vice versa. That surge of rising energy and growth in Hexagram 19 might be choked off if the orders fail to resonate with the army’s natural strength.

I’m far from confident about translating that into specific advice, for obvious reasons, but I would probably say something about checking and re-checking exact knowledge of capacity and demand – ensuring communication is perfect and instantaneous. Are there ‘rules’ he’s following that don’t match the situation and actually get in the way?

Also, the pictures made by the trigrams gets my attention: two versions of water stored in the earth, with the open, accepting quality of earth of supreme importance, and the first line becoming solid as the lake fills up. It brings to mind the image of pouring liquid into a vessel, filling all the spaces from the bottom up. If you pour everything in together, you can accommodate more. Do clients share cold storage space? Can they? I have – of course – no idea, but I would share the image with A. to see whether it gave him any ideas.

Here’s what the AI had to say. (You can read it without creating an account.)

How did it do, do you think? As you can see, I didn’t give YiShi the reading in the format it needed to start with, so that it began to cast a new reading. And also, I needed to tell it which line was changing: it doesn’t know that this is something it could work out from the hexagrams. Now I come to think about it, there’s no reason why it should – I’d been bamboozled because it appears intelligent.

It had the same basic idea of Hexagram 7 as I did. What it made of the moving line and relating hexagram was different, mainly because it’s using a different translation: it hasn’t heard of pitchpipes, and Wilhelm’s translation doesn’t have the idea of orders ‘blocking strength’. The use it makes of the translation it’s working with is hard to fault, in general.

It’s a bit odd that can get from “Thus the superior man is inexhaustible in his intention to teach and without limits in enduring and protecting the people” in 19’s Image to “This imagery stresses the importance of foresight and the ability to guide and protect. Your actions now must consider both the present demands and the eventual limits of your resources.” Didn’t the text just say he should ideally be without limits? The AI answer is good (if slightly platitudinous) common sense, but doesn’t seem to be using the reading.

It also doesn’t seem to be using the trigrams themselves at all – probably it would if it were specifically asked to. (I think AI generally does much better if you know what to ask it.)

It didn’t think that Hexagram 19 might be going beyond the question. From all I’ve seen so far, that’s a significant gap in AI readings. They can do a good job of applying the imagery to the question, and whatever background information you provide, but they can’t conceive of the conversation going beyond that. I tested this out later with an imaginary reading – more on that in a future blog post.

It did come up with very specific, practical suggestions, far more than I could have done. ‘Partner with neighboring facilities to temporarily lease additional space,’ ‘Utilize portable cold storage units for immediate capacity expansion’ – not anything I could have dreamt of, though I do wonder what these ideas have to do with the reading.

Final thought for now

After A.’s reading and a few extra experiments, I don’t see all that much of an issue with the quality of AI interpretation. It comes up with some good sense in its hexagram summaries, and some genuinely good ideas as it interprets imagery.

The problem is that A., as a relative newcomer, is left thinking that he’s done an I Ching reading. What the AI provides is a lot like reading the commentary, except that now you have commentary that talks directly about your question. And I realise – belatedly – that the problem with relying on the commentary isn’t primarily that it’s not applicable to your specific situation, but that you haven’t interacted with the oracle.

An I Ching reading is what happens when someone recognises what the oracle is saying to them. There’s some sort of breakthrough: escaping the limits of your usual way of seeing the situation, being carried well and truly outside the box. You engage emotionally; you become more fully present; you go from confusion to the sense of connection.

And if you don’t engage with the text yourself, none of that happens. There’s no moment of recognition, no emotional connection – no reading. Poor A. had looked at two AI interpretations, noted that they gave different advice, and wondered which was right. How was he to know? He had nothing to go on.

Of course, this is not a new problem, and not unique to AI. On the contrary, it happens when people rely on commentary, and it can happen with human interpreters, too:

‘What should I do?’
‘You should do this; the I Ching says so.’

The querent might think they’ve encountered the oracle, when really nothing’s happened at all.

I think AI can make this more of an issue, just because of its speed: it can produce great reams of in-depth interpretation, persuasively written, tailored to the specifics of the question and details of the background, within seconds, again and again. Given all this to digest, complete with headings and a neatly bullet-pointed summary at the end, it would be quite hard to see that you haven’t even begun the reading.

There’s more to explore here, so I’ll follow up later with another couple of posts. I asked Yi, ‘What does interpreting readings with AI provide?’ and the answer was Hexagram 26, Great Tending, changing at lines 1, 4 and 6 to 32, Lasting. It’s not hard to come up with ideas about 26.1 – ! – but the rest of the reading needs some thought. I tried feeding it to a few different AIs, too, to see what they’d come up with.

I also dreamt up an imaginary reading that I think painfully exposes the limits of AI – which, again, are not where I thought they would be.

More later!

I Ching Community discussion