There's a wonderful (British) book about this exact sort of thing only with people not AIs all doing pointless tasks: https://en.wikipedia.org/wiki/...

Companies love to *announce* that they will accept crypto at some point in the future, only to quietly either kill the project or to walk back the idea. Biggest and most obvious example, Tesla (https://apnews.com/article/tesla-buys-billion-bitcoin-061817c6795e75d1c3c9e9d6cfc4a911, "soon" in 2021)

But even the EFF walked back its Bitcoin donations back in 2011 (https://www.eff.org/deeplinks/2011/06/eff-and-bitcoin).

There are many others. It makes a good PR but is hard to operationally implement from a treasury and payments processing point of view. So I don't see how this is any different, especially given that airlines are extremely focused on cash management due to their razor-thin margins.

I posted below and I forgot to mention this use case - it's great for learning just about anything. It's infinitely patient and answers any stupid question I might have. It also very readily identifies when I'm outright wrong about something and corrects me.

I've surprised myself with some of the things I've been able to do with AI, particularly in the last few months. There are definitely limits - I'll give examples below - but from the mostly fun projects I've done, it's been quite enlightening. I've used both Ollama and the public AI's, mostly Gemini.

1. Code a daily random Wordle word email. This has been really helpful. I run a few "Wordle groups" where we all start with the same word. Before I had to use a website that picked a random word which was usually nonsense and I had to refresh until I got something useful. I had a local Ollama AI write me some database-driven Python scripts, one that scrapes the past words and stores them in the database and another that picks a random word from those remaining and sends a nicely formatted HTML email to me that I can share with the groups. Saves me a ton of time although there was a lot of testing and rework until it got it right. I didn't write a single line of code, on purpose. It's been running for over 6 months every day with no faults.

2. Build a simple circuit. I had Gemini build me a circuit with a power regulator, transistors, resistors, LEDs and a very specific fluid level sensor. It worked first try. What really surprised me was the questions it asked before it started building the actual circuit - it didn't make assumptions.

3. Build a test jig circuit. Gemini used an Arduino and designed a test jig for some hardware I wanted to test, with pull-up resistors and all the code required to display on a simple 16 character LCD and read the input from the device itself. The big problem with this project was that it used a lookup table for about 100 values and Gemini simply couldn't keep those numbers straight. I had to integrate that myself.

4. Write an AI torture chamber (for itself!) based on this https://youtu.be/7fNYj0EXxMs?s... crazy video. I was intrigued by this artist's vision so I had Ollama write code to basically have an AI be born, grow up learning from real news stories, get sick and eventually die leaving a final deathbed summary of its life. Dark stuff I know, but really fascinating. I used a model which had its limits removed so it was free to do anything. The nutty thing was that the "coding AI" gave lots of tips on how to make the "living AI" suffer and was totally on board with the whole thing. I am slightly concerned about that last part.

5. Write a screenplay. I loved The Black Hole (campy, Disney Sci Fi) as a kid and I wondered what a prequel to it would have been like. I wrote out what I thought the plot outline would be and then had Gemini write me an entire screenplay. It wasn't bad, certainly no worse than the original. Not gonna win any awards but now I know what that movie could have been about. Spoiler, Dr Hans Reinhardt is not a good guy.

6. Investigate an under-reported fraud. This my favorite use case. I used Gemini to gather sources, many really hard to find, about a crazy London-based fraudster who ran a FX trading firm. Then I fed those sources into NotebookLM and had it generate a research report and a podcast, the latter of which was actually compelling. This took a bit of work on my part.

7. Take a bunch of procedure documents from my business and look for weaknesses and potential regulatory issues. This worked OK but the funniest part was getting a podcast made from them. Literally the most boring podcast I've ever heard although I laughed quite a bit as the "hosts" tried to make this stuff sound interesting.

8. Making a podcast about my Dad's life. My Dad did Storyworth and we self-published his book at the end. I fed that book into NotebookLM again and got a really fun podcast made from it.

So yeah, these are actually helpful things. In addition, I've been using the public AIs to give me summaries of daily news stories, do search-engine type things, and occasionally get a deep dive on something I wrote to check it or just to hear a podcast about it. The circuit design stuff really surprised me, particularly where the Arduino coding intersected with it. But what I took away most from all these (and other) projects is that it makes a really good assistant that, while you don't give it free rein to do everything, if you give it enough "management", it can produce useful products - with limitations.

So yes, I do get use out of it.

Does the author mean to say that there will be quiet chaos, or do they mean to say that by making a quiet move, chaos will ensue?

Personally I like the concept of quiet chaos. Also would make a great metal band name.

Here's a quiz if you want to see your semicolon skills for yourself. Quite humbling.

https://style.mla.org/quiz-sem...

In some boroughs of London (Westminster for example) there are chargers on many if not most lamp-posts. Very cool.

I've always wondered about that sig. So I asked AI to write me a little story about it...

The annual "Fluffiest Pillow" competition was in chaos. Not the usual, genteel chaos of competitive pillow fluffing, mind you. This was⦠viscous.

Barnaby Thistlethwaite, a man whose mustache resembled a startled caterpillar, was weeping. His prize-winning goose-down pillow, "Cloud Nine," was now a gelatinous, trembling mass. Across the hall, Agnes Periwinkle, a woman who communicated exclusively through interpretive dance and kazoo solos, was attempting to sculpt her shredded memory foam pillow into a replica of the Eiffel Tower, with limited success.

The source of the pandemonium? Fisheye Fred.

Fred, a man whose left eye perpetually drifted towards the ceiling, had brought his âoerevolutionaryâ pillow-fluffing machine. It resembled a chrome-plated octopus wrestling a lawnmower. Heâ(TM)d proudly declared it would âoerevolutionize the very fabric of the pillow-fluffing paradigm!â

Heâ(TM)d plugged it in, pressed a large, ominous red button, and the machine had erupted in a symphony of grinding, whirring, and the distinct sound of something being violently pureed.

Now, feathers, stuffing, and the occasional errant spring floated through the air like morbid confetti. The walls were splattered with a disconcerting, lumpy paste. The judges, a panel of bewildered squirrels wearing tiny bow ties, had retreated to the ventilation system.

Barnaby, his mustache quivering, pointed a trembling finger at the machine, now emitting a faint, gurgling sound. "Fred," he whimpered, "what...what have you done?"

Agnes, mid-kazoo solo, paused, her eyebrows forming a question mark. She pointed at the ruined pillows, then at Fred, then mimed a blender.

Fred blinked, his good eye focusing on the chaotic scene. He scratched his head, dislodging a stray feather that floated lazily to the ground. "Oh," he said, his voice a low rumble. "You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"

Nod as as a wink to a blind man

Say no more

It still does in aviation. I just filled up with 100 "low lead" aviation fuel yesterday. The commercial business case for unleaded aviation fuel has never been made, hence it has never been produced in quantity.

I run a software company which uses MS SQL Server and .NET Core for our app development. My lead developer had a problem with a SQL script which was stubbornly slow - he was querying a huge table with an index that really wasn't helping the performance. He created a prompt to ChatGPT to ask how to optimize the query and it came back - instantly - with a subquery approach that moved the filtering logic from the index into memory. It ran significantly faster.

I have a feeling that GPT had somehow scraped some obscure Stack Exchange posting to come up with this but of course there's no way to really know.

Both of us were simply amazed at how efficient this approach was and we're going to try to use AI on other troublesome queries going forward.

So, yes, AI can help, at least in our limited experience with it.

I seem to remember it being extraordinarily costly to use back in the day. Can anyone remember how much it cost?

UK resident here. I use a power company called Octopus and every couple of weeks, they send me an email saying "free power between 1-2 pm" or some other time. This isn't a joke or a scam, they actually explain it as having too much energy in the grid and they need to burn it off. Because I have a smart meter, they can simply not charge me for that time.

They also have a night-time EV charging tariff which is 7p/kWh as opposed to the 35p/kWh which is again enabled by the smart meter.

What makes me slightly surprised about this whole thing is that I had assumed that the whole grid was pretty smart. Turns out not to be the case.

PS/2 = "Personal System for Two?"

Sounds romantic!

This leads me to an observation that magic, AI and 1,000 people in India are all indistinguishable.