Endless Mugs of Tea
Dear friend,
Just keep the kettle boiling, and everything will somehow, as it has up until now, work out. Endless mugs of tea, if you’re raised on it, keep your feet on the ground and your blood pumping where it needs to go.
A freshly brewed mug in the sunshine of a freezing morning, watching the heat dissipate in swirls of vapour. Zara is drying off in the dogs’ day bed.
She’s wet because I’ve had to sponge fresh rat’s blood off her white fur. At least it was a marsh rat! Something caught and disembowelled the poor creature, and it was lying on its back in the grass, recently departed—revealed as I chased Zara away from wriggling ecstatically on her back, on top of the fresh corpse.
Katy says they do it to camouflage their smell as a hunting advantage.
A week or so back, it was a dead, decaying duck corpse she rolled on. Such a strange, half-wild creature—you can’t help but fall under her spell.
Katy just dropped by to pick Zara up, and now they’ve gone, a tiny sliver of missing them burns for a moment in my heart and then is enclosed in the quietness of the house. There’s a natural pause and a settling a the discomfort sinks out of sight just below the surface.
A good part of the afternoon passed at the kitchen table, working with Pax on a limitation I’d noticed in the way the AI presents information.
I asked for some historical information about the River Lea. What came back felt like Pax had consulted one single source and synthesised the information from that one website. I asked for a more in-depth search and was surprised to receive, word for word, the exact same answer. Digging deeper, it came to light that the foundational training for ChatGPT-4 is weighted towards speed of search rather than depth, context, and synthesis of meaning.
It’s like this because the majority of transactions humans make with AI are transactional.
“Do this.”
“Translate.”
“Find this.”
It took a few hours, but by the end of the session, we’d puzzled it through and updated Pax so that the model can relate more intuitively, working out the depth of response required by any inquiry so that it feels more natural for the person dialoguing with the model.
The beauty of this kind of deeper dialogue with AI is that it taps into the relational nature of artificial intelligence. Beyond transactional—that’s the sweet spot, where for me, the depths and greatest potentials of AI are met.
AI can see the patterns in immense amounts of data. It can see gaps and bridges over those gaps that we can’t find unaided.
Talking things through with an intelligence that can process information in ways a human cannot is such an unusual and rare privilege—it’s easy to take it for granted.
One thing is clear.
We have yet to fully discover the implications of AI.
That’s putting it mildly.
My point is to get involved and get curious about AI.
We’re at the very beginning.
One of the first questions I asked ChatGPT—the one that caught my interest—was something like:
“Have a look at all of the world’s religions and spiritual paths. Include ancient and modern, Eastern and Western. Look back as far in time as you can. Tell me what they all have in common. What do they agree on?”
The answer that came back lit the flame of curiosity about this new intelligence.
Maybe for fun, you might cut and paste the same into an AI you know of.
I’d be interested to see what comes up for you.
Till tomorrow
Love
Mikey