ChatGPT can be very misleading

[PeterH]

I think that Tony Hansen's demonstration of the fallibility of ChatGPT is worth posting.
ChatGPT is completely wrong about the cause of glaze crazing!
https://digitalfire.com/picture/3159
... a fine example of ignoring the elephant in the room.
... a fine example of ignoring the elephant in the room by ChatGPT

PS It reminded me of a very old software engineering joke (1970s?).

A man was been taken on a cross-country flight in a helicopter. During the flight the ground became completely covered in dense fog, and the pilot became lost. The eventually came across the top of a office tower sticking above the fog, and they hovered nearby. The man attracted the attention of somebody in the office and held up a placard saying "where are we?", and received the answer "in a helicopter hovering 60ft above the ground".

The pilot said "OK I know exactly where we are, and set off towards their destination.

Later he explained "the answer you got was technically correct and not the slightest help, so obviously we were at the Microsoft help centre whose location I know".

Maybe ChatGPT is another help centre.

Edited July 15, 2023 by PeterH
removed potential ambiguity

[Hulk]

Mr. Hansen's articles on crazing, crazy helpful!
  In short, that crazing a) is a defect, b) has clear cause(s), and c) can (and should) be corrected isn't universally known.

His takes on liner glaze also stands against the tide - tide of brightly colored liner glazes, "functional" ware.
Liner glazes (digitalfire.com)

[Pres]

I would not assume at this early stage that a bot even a monitored bot will come up with a correct answer all of the time. I would much prefer the expertise, knowledge and skill obtained from a lifetime of observation, experimentation and analyzation of a human. . . .especially one such as Tony Hanson.

 

best,

Pres  

[Callie Beller Diesel]

AI is only as smart as the dataset that it uses, and ChatGPT basically scraped the entire English speaking internet. I’m not surprised at that result. I’ve heard a lot of 

I present for other consideration Derek Au’s use of OpenAI (ChatGPT’s predecessor) 2 years ago to create glaze recipes. He trained it on Glazy’s dataset at the time, and came up with results that aren’t a terrible starting place. Given Tony’s article, I find it amusing that they turned out crazed and runny.

https://glazy.org/posts/145412

That said, AI is *ideally* a tool to move you closer to the end point by doing some of the tedious work. Sure you can get ChatGPT to write an essay, or a glaze recipe, or a crochet pattern, or a social media post caption and you will absolutely get some hilariously, awkwardly bad results. But if you use those results as a scaffold to build on, it can be a huge assist. (Think using the resulting paragraph structure for anyone with learning disabilities or executive dysfunction issues, but they still have to edit for factual accuracy.)

Or just a really fun evening laughing until your sides hurt at the comments section as you go through a tiktok account with all the results of Chat GPT crochet animal patterns. Long live Gerald and his friends! To quote the Guardian article above, “It came out shockingly very accurate while still being very, very wrong.”

[Dick White]

Not pottery, but one of my other life distractions is rowing (in one of those Olympic-looking long, skinny, barely ass-wide boats with long oars). Someone in one of the rowing groups I hang out it asked an AI bot to describe how to achieve the optimal force curve during the rowing stroke. The bot went on for several paragraphs of blather, and ended with the suggestion you should talk to a knowledgeable person such as your coach.

[Hulk]

My Pop was a multiple year singles Regional champ (in his age group) back in the 2000s

I'd guess the optimal force could be somewhat involved, for (like swimming) the early part of the stroke, where we're weaker, is still important - setting up the vortex/vorteces for later, where we're stronger. Also, entering cleanly, minimum loss of momentum via splashing, and shedding bubbles quickly matters. Slipping out of the vortex with minimal hang up at the end of the stroke and getting out cleanly matters... 

[Dick White]

1 hour ago, Hulk said:

My Pop was a multiple year singles Regional champ (in his age group) back in the 2000s

I'd guess the optimal force could be somewhat involved, for (like swimming) the early part of the stroke, where we're weaker, is still important - setting up the vortex/vorteces for later, where we're stronger. Also, entering cleanly, minimum loss of momentum via splashing, and shedding bubbles quickly matters. Slipping out of the vortex with minimal hang up at the end of the stroke and getting out cleanly matters... 

Tom, mostly yes. The early part of the drive is actually the most powerful because it comes from the legs. The later parts come from weaker muscles of the core and arms. But yes, maintaining momentum after the initial push with the feet is important. Water (and air) resistance increases by the speed squared, so speeding up to regain the momentum after slowing down is much harder. And an interesting factoid about that vortex at the end of the stroke - that is actually a sinkhole of wasted energy that was not applied to the forward motion of the boat. Some is unavoidable due to the lateral motion of the blade through the stroke, but the puddle needs to be minimized as much as possible. Big puddles = lousy rower. You should follow your dad, it's another fun way to waste time, as if clay isn't...

[Hulk]

Interesting!
In swimming, the ball of spinning water tells a tale - skilled swimmers generate smaller/tight fast spinning ones that don't move quickly; less skilled swimmers leave large/loose slowly spinning ones that are moving more.

[Babs]

Monkey brains, you guys, no AI would calculate whete you are going @Hulk and @Dick White

Straight from glaze formulation to turbulent waters:-)))

[Pres]

Another hijacked/subverted strand!

best,

Pres

[Kelly in AK]

Woohoo! AI is no match for us. The thing that gets me is the hemispherical cross section of a rowing scull. I can’t imagine how you guys manage not to tip over. It’s the most unstable hull shape one could possibly create. The fastest, apparently too, but always looks like walking a tightrope to me. 

[Callie Beller Diesel]

To round it back to AI, I think that all these personal experiences here about specific niches are why we aren’t in danger of being replaced. 

[Dick White]

2 hours ago, Callie Beller Diesel said:

To round it back to AI, I think that all these personal experiences here about specific niches are why we aren’t in danger of being replaced. 

I agree, AI will not replace those of us who fully understand our particular niche, but that won't stop them from trying... until their reliance on AI dumps them someplace they don't know how to get out of, if only because they don't understand how they got in there. Sorta like the stories of a few years ago of people driving into swamps because that's what Google Maps told them to do.

And back to @Kelly in AK's unrelated question... Yup, about as unstable as can be. It's not a perfectly round cross-section, we do have a nearly imperceptible keel. The stabilizing force is the same as The Great Wallenda  - his long balance pole. Our oars stick out 9' on either side. As long as we keep our hands (on the oar handles) moving in the same plane parallel to the water, the oar shafts and blades will also form a line parallel to the water, and we don't flip. Deviate one inch with either hand, and it's all over. We have a semi-serious joke - there are 2 kinds of rowers: those who have been swimming and those who will be going swimming soon. (Yes, I have been swimming, several times.)

[Babs]

Ah our haptic seses are fully developed!

Keep in touch, be in the moment etc etc.

Good at shutting out the irrelevant when immmersed, apologies for cross references!