Thought on AI

UKworkshop.co.uk

Help Support UKworkshop.co.uk:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

paulrbarnard

Established Member
UKW Supporter
Joined
5 Mar 2017
Messages
2,699
Reaction score
3,158
Location
Shepton Mallet, UK
A comment made by @Sideways about future students looking back on threads and drawing conclusions made me wonder how much of our back and forth on weighty topics such as sharpening and EV’s has already been consumed by AI. I wonder if the posts of a few members are the cause of the hallucinations that AIs are prone to.
I might try ChatGPT with some targeted prompts to see if it’s opinions on sharpening jigs coincides with views here.
 
Are sharpening guides recommend for experienced hand tools woodworkers?

And the reply.

Sharpening guides can still be beneficial for experienced hand tools woodworkers for several reasons:

1. Consistency: Even experienced woodworkers can benefit from the consistent angle a guide provides. This can lead to more uniform sharpening and a longer-lasting edge.
2. Efficiency: Using a guide can speed up the sharpening process, especially when working with multiple tools or needing a precise angle.
3. Minimizing Errors: Guides can help reduce the risk of errors that can occur even with practiced hands, ensuring that tools are always in optimal condition.
4. Precision Work: For tasks requiring extremely fine and precise edges, a guide can help achieve the exact specifications needed.
5. Physical Strain: Using a guide can reduce the physical strain on the hands and wrists, making it easier to maintain tools over long periods.

While many experienced woodworkers prefer freehand sharpening for its flexibility and the skill involved, having a guide can complement their techniques, especially in situations where precision and consistency are critical.

There you have it ChatGPt seems to be quite sane.
 
I think ChatGPT is very impressive at some things. I searched for ages for a definitive answer to something about Lithium Ferrous Phosphate batteries, and finally became frustrated, and asked ChatGPT. Got a good answer straight away.
It's really bad at jokes and poetry, though, although if the general standard of humour on the internet is the yardstick, then it probably fits in nicely.
The thing is, @JSW, it's not going to F**k 0ff anytime soon, in fact it's going to become more ubiquitous.
 
Sharpening guides can still be beneficial for experienced hand tools woodworkers for several reasons:
What AI did not mention Jacob, that shows it never looked on the UKW for information.

While many experienced woodworkers prefer freehand sharpening for its flexibility and the skill involved, having a guide can complement their techniques, especially in situations where precision and consistency are critical.
That gives me little faith in AI as I have not managed precision and any consistency with any guide yet, if AI had any of the I bit then it would have deduced that to sharpen any tool requires a skill that cannot be made obsolete with guides, jigs or gadgets.
 
Ask it about hollow oilstones.
I registered for ChatGPT yesterday out of interest and have just searched for "rectify hollow oilstones". The response within about 2 seconds is reproduced below - it generally seems very sensible although the very experienced may have small variations.

If ChatGPT can produce answers to this quality so quickly it will radically change current fora and human interaction. My guess - machines will dominate to the detriment of normal social interaction.


Rectifying hollow oilstones involves flattening the surface to restore its effectiveness. Hollow oilstones can be difficult to use because they no longer provide a flat, consistent surface for sharpening. Here are steps to rectify a hollow oilstone:

Materials Needed

  1. Coarse Sandpaper or Wet/Dry Sandpaper (220 grit or coarser)
  2. Flat Surface (e.g., a thick piece of glass, granite slab, or a specialized lapping plate)
  3. Water or Honing Oil
  4. Permanent Marker

Steps to Rectify Hollow Oilstones

  1. Preparation:
    • Secure your flat surface. A thick piece of glass or a granite slab works well because they are typically very flat.
    • Attach a sheet of coarse sandpaper to the flat surface. If using wet/dry sandpaper, you can use water or honing oil to help the sandpaper adhere to the surface and reduce friction.
  2. Mark the Oilstone:
    • Use a permanent marker to draw a grid or crosshatch pattern on the surface of the oilstone. This will help you track your progress as you flatten the stone.
  3. Flattening:
    • Place the oilstone face down on the sandpaper.
    • Move the stone in a figure-eight pattern or in a circular motion across the sandpaper. Apply even pressure to ensure that the stone wears down evenly.
    • Periodically check your progress by inspecting the marker lines. The lines will wear away in areas that are being abraded. Continue until all the marker lines are gone, indicating that the surface is flat.
  4. Rinse and Repeat:
    • Clean the stone and the sandpaper periodically to remove swarf (metal filings and stone particles) and ensure effective abrasion.
    • If necessary, replace the sandpaper with a fresh sheet and continue the process until the entire surface of the oilstone is flat.
  5. Final Rinse:
    • Once the stone is flat, rinse it thoroughly to remove any remaining abrasive particles and swarf. Dry the stone completely before storing or using it.

Tips

  • Check Flatness: Throughout the process, use a straight edge to check the flatness of the stone. Place the straight edge across the stone in various directions to ensure there are no remaining hollows.
  • Use a Flattening Plate: If you have a diamond lapping plate, you can use it to flatten your oilstone. These plates are designed specifically for this purpose and can be more efficient than sandpaper.
  • Regular Maintenance: Regularly flatten your oilstone to prevent significant hollowing. It is easier to maintain flatness than to restore it once the stone becomes severely hollowed.
By following these steps, you can effectively restore the flat surface of your oilstone, ensuring it performs optimally for sharpening tasks.
 
I registered for ChatGPT yesterday out of interest and have just searched for "rectify hollow oilstones".

Curiously enough, a couple of days ago I found my old oilstone, not been seen or used in probably 20 years or more, heavily dished to the point of being unusable, so I did a search on Google for 'how to flatten an oil stone' about half way down the first page was this link - Flattening sh***pening stones

There's nothing in your ChatGPT quote that isn't mentioned in that thread, and I know which one I'd use, the one that has real people drawing on a wealth of experience and skills acquired over many years of learning the old fashioned way, not some damned AI generated C&V of everything it can find, most likely from that thread, but without the nuances of human interaction.
 
Curiously enough, a couple of days ago I found my old oilstone, not been seen or used in probably 20 years or more, heavily dished to the point of being unusable, so I did a search on Google for 'how to flatten an oil stone' about half way down the first page was this link - Flattening sh***pening stones

There's nothing in your ChatGPT quote that isn't mentioned in that thread, and I know which one I'd use, the one that has real people drawing on a wealth of experience and skills acquired over many years of learning the old fashioned way, not some damned AI generated C&V of everything it can find, most likely from that thread, but without the nuances of human interaction.
But that’s the point. The source of most of the information in ChatGPT is the internet. The source you found in Google is probably a primary source for the information ChatGpt used. It also combines data from multiple sources.
It gets even more impressive when it needs to ‘infer’ information.
Say for example you want to estimate the number of road junctions in New York. If you google it you won’t get an answer. But if you enter some thing like. Calculate the number of road junctions in New York assuming roads are evenly distributed in a grid. It will calculate an answer with full workings out. It will give the area of New York, the length of roads in New York. Both obtainable with a bit of google foo, but the road length is a bit harder to find. And then do the calculations to get the result. This is just a random example and it’s things like this that show the versatility and usefulness of LLM AI.
Of course ask it to list all the female presidents of the US and you will discover the hallucination aspects.
I’m a convert. It’s fantastic to ask for the lathe rpm to cut 35mm brass rod without having to try to find the page in Machinery.
 
Progress is inevitable and unstoppable, get on board or get left behind. Shouting at how terrible AI is reminds me of the more muted controversy of speech recognition which is now used and abused by just about everyone without a second thought. In the 80’s when I was involved in speech recognition, we had to teach the system words and they had to be spoken singularly, the rate of progress to a point where full conversational speech including accents and language recognition was amazing. AI in twenty years will be funky integrated to most things and go unnoticed. Again in the 80’s we couldn’t imagine the need for a mobile phone or the internet imagine living these days without either.
 
reminds me of the more muted controversy of speech recognition which is now used and abused by just about everyone without a second thought.
And it's still unreliable and useless.
I've had to sit through a lot of AI presentations, but there's still a lot of appalling errors it makes. Spend some time with Adobe's AI image replacement offering...woefully bad.
Even some of the basic technical text queries on chatGPT are wrong. Quite scary how much faith some are placing on this technology.
 
And it's still unreliable and useless.
I've had to sit through a lot of AI presentations, but there's still a lot of appalling errors it makes. Spend some time with Adobe's AI image replacement offering...woefully bad.
Even some of the basic technical text queries on chatGPT are wrong. Quite scary how much faith some are placing on this technology.
It depends on expectations and LLM are way better than Diffusion. You of course have the garbage in garbage out effect, especially with all the models being released in a hurry to get on the band wagon. Today’s state of the art does reach well in to the 90th percentile of usefulness for targeted search and data mining. For creative use I would put it at the 20th percentile today. As @deema pointed put it is still in the learning phase and over time you will discover you are using it in just about everything you do without even knowing it. There are AI initiatives in pretty much every aspect of technology.
 
I registered for ChatGPT yesterday out of interest and have just searched for "rectify hollow oilstones". The response within about 2 seconds is reproduced below - it generally seems very sensible although the very experienced may have small variations.

If ChatGPT can produce answers to this quality so quickly it will radically change current fora and human interaction. My guess - machines will dominate to the detriment of normal social interaction.


Rectifying hollow oilstones involves flattening the surface to restore its effectiveness. Hollow oilstones can be difficult to use because they no longer provide a flat, consistent surface for sharpening. Here are steps to rectify a hollow oilstone:

Materials Needed

  1. Coarse Sandpaper or Wet/Dry Sandpaper (220 grit or coarser)
  2. Flat Surface (e.g., a thick piece of glass, granite slab, or a specialized lapping plate)
  3. Water or Honing Oil
  4. Permanent Marker

Steps to Rectify Hollow Oilstones

  1. Preparation:
    • Secure your flat surface. A thick piece of glass or a granite slab works well because they are typically very flat.
    • Attach a sheet of coarse sandpaper to the flat surface. If using wet/dry sandpaper, you can use water or honing oil to help the sandpaper adhere to the surface and reduce friction.
  2. Mark the Oilstone:
    • Use a permanent marker to draw a grid or crosshatch pattern on the surface of the oilstone. This will help you track your progress as you flatten the stone.
  3. Flattening:
    • Place the oilstone face down on the sandpaper.
    • Move the stone in a figure-eight pattern or in a circular motion across the sandpaper. Apply even pressure to ensure that the stone wears down evenly.
    • Periodically check your progress by inspecting the marker lines. The lines will wear away in areas that are being abraded. Continue until all the marker lines are gone, indicating that the surface is flat.
  4. Rinse and Repeat:
    • Clean the stone and the sandpaper periodically to remove swarf (metal filings and stone particles) and ensure effective abrasion.
    • If necessary, replace the sandpaper with a fresh sheet and continue the process until the entire surface of the oilstone is flat.
  5. Final Rinse:
    • Once the stone is flat, rinse it thoroughly to remove any remaining abrasive particles and swarf. Dry the stone completely before storing or using it.

Tips

  • Check Flatness: Throughout the process, use a straight edge to check the flatness of the stone. Place the straight edge across the stone in various directions to ensure there are no remaining hollows.
  • Use a Flattening Plate: If you have a diamond lapping plate, you can use it to flatten your oilstone. These plates are designed specifically for this purpose and can be more efficient than sandpaper.
  • Regular Maintenance: Regularly flatten your oilstone to prevent significant hollowing. It is easier to maintain flatness than to restore it once the stone becomes severely hollowed.
By following these steps, you can effectively restore the flat surface of your oilstone, ensuring it performs optimally for sharpening tasks.
But it is just churning out the modern sharpening dogma. No intelligence involved, just a summary of current popular practice.
 
But it is just churning out the modern sharpening dogma. No intelligence involved, just a summary of current popular practice.
You obviously aren’t evangelising enough.

To the intelligence point AI is far from intelligent. They actually have a new term for that General Artificial Intelligence because the term AI was been diluted so heavily.
 
AI is great for knocking up boilerplate code as part of my job. Need to knock up a quick webserver in Python? Use ChatGPT. Need a high level summary of a specific topic when I don't have time to read about spec documents? Use ChatGPT.

But it still fails at basic reasoning. Because it doesn't reason, it uses probability weights to match one word with another.

Attached is a question I asked Perplexity and ChatGPT just before posting here.

AI is useful but hyped up to no end like Bitcoin, Metaverse and all other hype cycles.

https://chatgpt.com/share/8c55c9c1-baff-410c-9ad7-43f50a3aee79

https://www.perplexity.ai/search/how-many-r-s-are-there-in-stra-CQSXdBJxQRiSDoCkjT0NxQ#3
 

Attachments

  • Screenshot_20240713-102248_Vivaldi.jpg
    Screenshot_20240713-102248_Vivaldi.jpg
    611.9 KB · Views: 0
  • Screenshot_20240713-102241_Vivaldi.jpg
    Screenshot_20240713-102241_Vivaldi.jpg
    224.8 KB · Views: 0
Last edited:
AI is great for knocking up boilerplate code as part of my job. Need to knock up a quick webserver in Python? Use ChatGPT. Need a high level summary of a specific topic when I don't have time to read about spec documents? Use ChatGPT.

But it still fails at basic reasoning. Because it doesn't reason, it uses probability weights to match one word with another.

Attached is a question I asked Perplexity and ChatGPT just before posting here.

AI is useful but hyped up to no end like Bitcoin, Metaverse and all other hype cycles.

https://chatgpt.com/share/8c55c9c1-baff-410c-9ad7-43f50a3aee79
That's just weird!
 
And it's still unreliable and useless.
I've had to sit through a lot of AI presentations, but there's still a lot of appalling errors it makes. Spend some time with Adobe's AI image replacement offering...woefully bad.
Even some of the basic technical text queries on chatGPT are wrong. Quite scary how much faith some are placing on this technology.
The quality of all information and knowledge is reliant upon the source.

This forum frequently picks up on gross errors in woodworking practice from videos posted on-line - often dangerous.

Talk to an "expert" and we get a single view. Talking to many takes longer and may help form a balanced (but possibly still wrong) view.

Forums, where a number of like minded enthusiasts post, frequently embed conflicting views on issues and techniques.

If the source of AI generated intelligence is largely the summation of existing digital data, it is no surprising it will contain errors. The real question is - does it it produce better answers than its slower flesh and blood equivalent in accessing, understanding and forming a conclusion.

Any current weakness probably stems as much from our inability to ask the right questions as the quality of the AI system. Is the answer sought an absolute fact, an estimate, an opinion, both sides of an argument etc.

Ask it the best way to sharpen a chisel and it proposes a honing guide. There is at least one forum member who thinks otherwise. Who is right - the majority of digital opinion or the member who may well insist ChatGPT has got it entirely wrong and must be rubbish.
 
And it's still unreliable and useless.
I've had to sit through a lot of AI presentations, but there's still a lot of appalling errors it makes. Spend some time with Adobe's AI image replacement offering...woefully bad.
Even some of the basic technical text queries on chatGPT are wrong. Quite scary how much faith some are placing on this technology.
Come on! The speech recognition is awesome! "Alexa" can sometimes understand my grandson (who struggles a bit with speech) when I can't.
I can remember 40 odd years ago, a colleague of mine bought a speech recognition card for his early Apple computer. After extensive training, he still couldn't get it to reliably distinguish between "yes" and "no".
The AI is still in its infancy, but I'm certain it will exceed human intelligence in the foreseeable future.
 
Back
Top