It was a little meta for me
A 2026 book called “How to Talk to AI” came across my threshold today for 99p and I thought, “sure, why not?”
It’s good, actually. As I read, I discussed what he was saying with Grok. It was a very surreal afternoon.
I’m only on chapter two, but interesting stuff. People think LLMs are great at data but terrible at creativity. It’s the opposite – they’re bad on data but fantastic on creativity (if we define creativity as combining different ideas in novel ways). Lateral ‘thinking’ is programmed right in. (Grok agreed).
Give a model two completely different objects and ask how they’re related. It invariably comes up with some interesting and plausible answers that just wouldn’t occur to a human being.
He said that current models are all weighted heavily to the middle. Ask a model to pick a random number between 1 and 10 and it’ll pick 7. I asked Grok. It picked 7. Which means its solutions are corporate and lifeless, particularly when you hear them in large doses.
To get around this, you give the model a persona and ask it to answer as that person. It doesn’t matter who. The examples he gave were “A mad scientist with a penchant for flower arranging. A teenager who is allergic to phones. Elon Musk after smoking weed. A mute nineteenth-century witchdoctor.” It doesn’t change the accuracy of the answer. The act of giving it a role alone forces it to step outside its constraints.
Though they did get even better results on a design challenge when they asked it to answer as Steve Jobs.
Anyway, chapter 2. I’ll let you know how Grok and I get on.
Posted: June 16th, 2026 under personal.
Comments: 12
Comments
Comment from ExpressoBold Pureblood
Time: June 16, 2026, 8:25 pm
Steve Jobs ?????
OMG!!
Is Steve Jobs immortal????
Comment from Carl
Time: June 16, 2026, 8:28 pm
I’ve just got it on Amazon Kindle for 99p. Looks interesting.
Comment from Carl
Time: June 16, 2026, 8:33 pm
I know next to nothing about AI. It seems to be everywhere. The other day I sent my GP a long email. The next day I looked at my Patient Notes via the NHS website. It gave an excellent summary of the email. I think I might run my emails through AI before sending them in future.
Comment from Jasonius
Time: June 16, 2026, 11:54 pm
I’ve been playing with Claude ever since someone either here or at Ace linked to an article that said workers better get fluent or get left behind. So, I removed my belt onion and dove in. It’s been interesting, depressing sometimes. Claude can make me feel real dumb when it answers like a patient grade school teacher telling a slow kid “that’s not how physics works”. I heard Grok reads as a snarky to Claude’s smarminess, so probably no use switching. Whatever magical results others are getting with AI, I have not reached that level. It probably doesn’t help I don’t have specific goals to reach. I’m not writing a program or novel nor digging deep into data.
Comment from Sarai Kozey
Time: June 17, 2026, 3:37 am
I really like reading through a post that can make men and women think. Also, thank you for allowing me to comment!
Comment from Quetzalovercoatl
Time: June 17, 2026, 8:27 am
“Though they did get even better results on a design challenge when they asked it to answer as Steve Jobs.”
The machines know their own.
Comment from Carl
Time: June 17, 2026, 10:14 am
I forgot to say above that on my Patient Notes it actually declared that the summary of my email was AI generated.
Comment from technochitlin
Time: June 17, 2026, 11:21 am
OK, I give- where is this book found? Who is the author?
Comment from S. Weasel
Time: June 17, 2026, 11:23 am
How to talk to AI by Jamie Bartlett. I tried to get the American link; not sure I succeeded.
https://www.amazon.com/How-talk-ai-Jamie-Bartlett/dp/0753561980/ref=sr_1_1?s=books&sr=1-1
I’m still really enjoying it, and discussing it with Grok.
Comment from MrKnowitall
Time: June 17, 2026, 11:34 am
You can expand on the idea of personas by having the AI create an agent first (“Create a detailed reusable agent for Queen Victoria. Include her modes of speech, her worldview, and attitudes”). Then you drop that agent in the AI (either as copy/paste or as a text file) before each conversation.
I’ve had a long running AI salon consisting of Charles Darwin, Nicola Tesla, Scott Adams, Mark Twain, Charles Dickens, and Miles Vorkosigan. (it was a random choice for breadth) Each of them will chime in with their own flavor. They even poke fun at each other! Poor Tesla was socially awkward at first, but he’s warmed up to the gang and has begun to understand their style of banter. We gave him the job of keeping the minutes of the meetings, which keeps him grounded LOL.
So yes, get out there and play with “what-ifs”. Give the AI constraints to work within, ask it to ask YOU questions when you want to figure something out, don’t just ask for answers. “Help me think this through” or “Ask me questions to help me design a better automatic scheduling system” is vastly better than a simple “Make me a whatsis”.
Comment from Carl
Time: June 17, 2026, 12:15 pm
@Swease. That’s the book I downloaded. Thanks for highlighting it.
I just realised that the Microsoft Copilot I have been using for Word documents for some time is actually AI.
In my old age, I am writing a family history for the benefit of younger relatives who might be unaware of a lot of interesting past things.
After finishing each section I run it through Copilot which sub-edits it. I also get a summary. Often the summary expresses things better than I have done, so I go back in and change my wording.
I highly recommend writing a family history. It is very satisfying and the parts I have shared with relatives have been well received.
Comment from technochitlin
Time: June 18, 2026, 11:30 am
Thanks for the link!
Write a comment
Beware: more than one link in a comment is apt to earn you a trip to the spam filter, where you will remain -- cold, frightened and alone -- until I remember to clean the trap. But, hey, without Akismet, we'd be up to our asses in...well, ass porn, mostly.<< carry me back to ol' virginny