Comments on Lift Up Your Hearts!

Comment on Copilot Lied to Me

Stephan — Fri, 17 Apr 2026 01:08:11 -0400

My understanding of AI—which may be wrong and outdated—is that it’s basically trying to guess the correct answer, kind of like a high school kid that didn’t read the assignment. That’s what it’s doing with the sheet of handwriting, that’s what it’s doing with the imaginary second sheet, that’s what it’s doing when apologizing.

Whether that’s a moral failing, and whose it might be, is an interesting question.

Comment on Copilot Lied to Me

Joyful — Wed, 15 Apr 2026 07:21:06 -0400

First, I'd like to point out two transcription differences I found.
There are two sentences ending with "Bryson City." The first ends a paragraph. The second starts the new paragraph. And then the next sentence says "We [made] almost daily trips..."

I didn't read through all of Grandpa's handwritten page, but that's what I found in the part I did read.

Second, my conclusion to your main point is that AI is not given moral guidelines at all. It knows how to apologize when someone mentions that they don't like what it says, but it has no underlying moral framework.

It may have a form of intelligence, knowledge of information, but it is not human.

Comment on AI Photo Editing Success

Beth — Mon, 13 Apr 2026 22:45:19 -0400

WOW!!!

Comment on LLM Transcription of Printed Text

Grace K — Thu, 09 Apr 2026 10:14:46 -0400

Impressive. Susie also loves ChatGPT, but I remain WAY leery of AI, despite its obvious advantages.

Comment on LLM Transcription of Printed Text

SursumCorda — Thu, 09 Apr 2026 08:28:49 -0400

And wait till you see my upcoming post on what it did for me with text that was almost totally obscured by bleed-through from the other side of the paper.

Comment on LLM Transcription of Printed Text

SursumCorda — Thu, 09 Apr 2026 08:26:27 -0400

I have not yet started working on handwritten text, but yesterday I tried a random page from Grandpa's handwritten journals, and was very impressed with the result. His handwriting is not easy for me to read, even knowing him and the context. ChatGPT's transcription wasn't perfect—for example it transcribed '3 lengths of 2 1/2" hose and 2 lengths of 1 1/2" hose' as '3 lengths of 2×4’s here and 2 lengths of 1×? here'. A tough one!

It's not something you can just throw text into without expecting to proofread and make corrections. But overall its performance was remarkable: far, far better than anything else I've seen so far. And I'm only using the free version of ChatGPT. I'm looking forward to doing more with it.

Comment on LLM Transcription of Printed Text

Joyful — Thu, 09 Apr 2026 08:03:27 -0400

Nice! Have you tried it on handwritten pages? I just took a photo of yesterday's journal entry and asked Gemini to transcribe it. It was reasonable, mostly accurate, but it did make several mistakes. And yesterday my handwriting was pretty good. (:

Comment on Doing the Lambeth Walk

SursumCorda — Wed, 01 Apr 2026 10:16:39 -0400

Whenever I hear "Lambeth," I think of Lambeth Palace, the residence of the Archbishop of Canterbury. Do you suppose that someone in that position, in an unguarded or frivolous moment, ever did the Lambeth Walk?

Comment on Using AI for Mystery Photos

Grace K — Sat, 28 Mar 2026 10:19:18 -0400

You have added to my already negative view of AI. Admit I'm an old curmudgeon who enjoys some AI benefits while terrified of what the future will bring.

Comment on The Remarkable Cousin Sylvia

Grace K — Sun, 22 Mar 2026 12:22:41 -0400

Love these unexpected snippets of family history.

Comment on Using AI for Mystery Photos

Stephan — Mon, 16 Mar 2026 14:48:38 -0400

I’ve yet get copilot to correctly draw a wireframe cube with a body diagonal. It has always started out at one corner and, instead of connecting it to the opposite corner, drawn a line to roughly the middle of one of the most distant edges. Attempts to correct it lead to copilot drawing the same image while insisting it’s drawing corner to corner.
But it’s surprisingly good at drawing the diagonal in front of and behind the appropriate edges.

Comment on How Much We Unintentionally Reveal to Eyes That Can See!

Laurie — Sat, 14 Mar 2026 08:11:00 -0400

Oh. My. Fascinating. So many aspects are accurate; "it" seems to be far more attentive and perceptive than most people are... which is troubling - a caution to use "it" carefully and to sift! Thank you for this; I am intrigued.

Comment on Just Do Your Part

SursumCorda — Wed, 04 Mar 2026 07:44:42 -0500

Based on what he says, I imagine he would regard that as individual and personal.

Comment on Just Do Your Part

Stephan — Wed, 04 Mar 2026 07:24:24 -0500

I wonder what he means by “do your part?”

Comment on There Was a Super Bowl?

Dad-o — Wed, 18 Feb 2026 04:35:57 -0500

Or watch a lecture on quantum mechanics by the Great Courses, which we did the last time the Patriots actually won a Super Bowl.