-
I ask the chatbot for fun the steps needed to refret a 1945 Gibson L5 and what tools would be needed. I tell you surprisingly it was very good and basically, I agreed with everything it said except the fret size to use. It provided details on the points need to address. Frankly it could be a lot worse, I am impressed.
Refretting a vintage guitar like a 1945 Gibson L5 requires skill, precision, and care to preserve its value and playability. Here's a detailed guide on the steps and tools you'll need:
Preparation- Assess the Fretboard and Frets
- Check the fret height, tang depth, and wear.
- Measure the fret dimensions with a digital caliper to match replacements closely.
- Note the fret tang width and depth since Gibson guitars of that era often used smaller tangs.
- Choose Replacement Frets
- Select fret wire that matches the original size or choose a size that suits your playing style. For vintage feel, medium or medium-jumbo frets are common.
- Tools Needed
- Fret Puller: Specialized pliers to remove frets without damaging the fretboard.
- Heat Source: A soldering iron to heat frets for easy removal (reduces risk of chipping the wood).
- Radius Gauge: To check the fretboard radius.
- Fret Hammer or Press: For installing new frets.
- Fret Bender: To match the fret wire radius to the fretboard.
- Files: For beveling, shaping, and smoothing fret ends.
- Fret Cutter: To trim excess fret wire.
- Masking Tape: To protect the fretboard and finish.
- Fret Leveling File or Beam: For leveling frets after installation.
- Crowning File: To reshape frets after leveling.
- Sandpaper/Steel Wool: For final polishing (e.g., 320–800 grit, then 0000 steel wool).
- Glue (Optional): Thin super glue or wood glue if required.
- Vacuum or Magnet: To clean metal filings.
- Light/Magnifying Glass: For precision work.
- Straightedge: To check fretboard and neck straightness.
- Notched Straightedge: To check fretboard relief.
Step-by-Step Refret Process- Secure the Guitar
- Place the guitar on a neck rest or work mat. Ensure it is stable and well-supported.
- Remove Old Frets
- Heat each fret with a soldering iron for 5–10 seconds to soften any glue.
- Gently lift the frets using a fret puller. Work slowly to avoid chipping.
- Prepare the Fretboard
- Sand the fretboard lightly with fine sandpaper (320–400 grit) to remove old glue or debris.
- Check for and fix any chips using wood dust mixed with super glue.
- Clean the fretboard and check the radius with a radius gauge.
- Cut and Shape the New Frets
- Use a fret bender to match the fretboard radius.
- Cut each fret to length using a fret cutter, leaving a bit of overhang.
- Install the New Frets
- Align the new fret into the slot. Tap it gently with a fret hammer or use a fret press for even pressure.
- Use glue sparingly if necessary, especially if the slots are loose.
- Trim excess fret ends with a fret cutter.
- Level the Frets
- Use a leveling file or beam to ensure all frets are even. Mask the fretboard with tape to protect it.
- Mark the fret tops with a marker and file until the marks are evenly removed.
- Crown the Frets
- Use a crowning file to reshape the fret tops into a rounded profile.
- Smooth the Fret Ends
- File and bevel fret ends at a 30–35° angle. Smooth sharp edges with a fine file.
- Sand the fret ends with progressively finer sandpaper and finish with steel wool.
- Polish the Frets
- Polish the frets with 800–1200 grit sandpaper or micro-mesh and finish with steel wool or a polishing compound for a mirror-like shine.
- Final Setup
- Check and adjust the neck relief and string action.
- Reinstall strings and ensure no buzzing or uneven frets.
Tips for Working on a Vintage Guitar- Preserve Originality: Avoid over-sanding the fretboard or modifying the slots unless absolutely necessary.
- Keep the Binding Intact: If the guitar has bound edges, ensure the frets are trimmed and finished cleanly at the binding.
- Test on Scrap: If you're unfamiliar with certain tools, practice on a scrap piece before working on the guitar.
- Document the Process: Take photos to document the condition of the guitar before and after the refret.
Would you like me to help you find specific fret wire dimensions or tools for your Gibson L5?
4o
- Assess the Fretboard and Frets
-
12-27-2024 06:45 PM
-
Tbf this is all beyond my field of knowledge. I do remember attending a lecture by an “AI” specialist who was quick to identify the failings of LLMs which accord with what other knowledgable people in the area have said.
Originally Posted by djg
Correct me if I’m wrong but as I understand it Hopfield’s work got us to where we are with LLMs?
Nobel laureates themselves are of note because of their Nobel winning work (and of course other important work they may have done).
They themselves are human people and can sometimes believe some nutty stuff. Take Kary Mullis for an extreme example. James Watson for a nasty one. Not sure about Penrose. I like his ideas on consciousness but they are not mainstream.
OTOH we are subjected to a hosepipe of hype from talky people on podcasts and the like who often have a vested financial interest in the success of the sector. These include the CEO’s of tech companies whose first responsibility is of course not to accurately represent the capabilities of their products but to enhance shareholder value.
There’s two possible reasons why for example Musk would simultaneously warn of the dangers of unfettered AI while continuing to … pour money into researching AI, and this is the most banal and sensible one. Hype is hype and it’s in his self interest to feed into the general buzz around the field. That’s Musk’s trade, the attention economy. OTOH the tech world is built on continuous growth. There’s a lot of watchers saying that the low hanging fruit is gone and this is heading into bubble territory. There aren’t endless more new users. Moore’s law fell off. The internet and computing is becoming mature.
Are they right? Don’t know.
But remember the metaverse? That was a punt from Zuck that didn’t pay off. Maybe quantum computing next year. AGI LLMs powered by AI designed Stellarator fusion! Bootstrap the singularity!!! Download Cheese!!!!
They are all looking for something to be the next big thing, and Ai is it right now. It’s being inserted everywhere whether it’s appropriate or not. The answer to all questions. Google search. Great idea. I’ll go and eat my daily rocks now.
But it’s not for us. It enhances shareholder value.
OTOH I don’t mean to dunk on LLMs. They and the engineers who developed them are not to blame for any of this. And there some things LLMs are amazing at - they seem to have a lot of potential in medicine and biotech. Could do a lot of good for humanity!
Sent from my iPhone using TapatalkLast edited by Christian Miller; 12-27-2024 at 07:29 PM.
-
I asked what are the best Gibson jazz guitars of all time
Gibson is renowned for its jazz guitars, especially its archtop models, which have been staples of jazz guitarists for decades. Here’s a list of some of the greatest Gibson jazz guitars of all time:
1. Gibson L-5
• Overview: Introduced in 1922, the Gibson L-5 is often considered the ultimate jazz guitar. It was the first guitar to feature F-holes, giving it a distinct look and a bright, articulate tone.
• Notable Players: Wes Montgomery, Eddie Lang, and Johnny Smith.
• Why It’s Great: Warm, rich tone with exceptional craftsmanship and timeless aesthetics.
2. Gibson ES-175
• Overview: Released in 1949, the ES-175 is one of the most iconic electric archtop jazz guitars, thanks to its affordability and versatility.
• Notable Players: Joe Pass, Herb Ellis, and Pat Metheny.
• Why It’s Great: Its PAF or P-90 pickups deliver a classic jazz tone with a touch of warmth and clarity.
3. Gibson Super 400
• Overview: Launched in 1934, the Super 400 is Gibson’s largest and most luxurious archtop guitar.
• Notable Players: Scotty Moore, Kenny Burrell.
• Why It’s Great: Its size and craftsmanship result in unparalleled projection and tonal richness.
4. Gibson ES-335
• Overview: Introduced in 1958, the ES-335 is a semi-hollow guitar that blends the warm tones of an archtop with the sustain and feedback resistance of a solid-body.
• Notable Players: Larry Carlton, John Scofield.
• Why It’s Great: Its versatility makes it suitable for jazz and other genres.
5. Gibson Byrdland
• Overview: Named after Billy Byrd and Hank Garland, this model was introduced in 1955 as a thinner, shorter-scale version of the Super 400.
• Notable Players: Ted Nugent (though more in rock contexts), Hank Garland.
• Why It’s Great: Combines playability with a smooth, rounded jazz tone.
6. Gibson ES-350T
• Overview: A thinner and more lightweight version of the ES-350, this guitar became popular among jazz players for its portability and tone.
• Notable Players: Chuck Berry (used in jazz and rock).
• Why It’s Great: Its smooth, articulate tone is perfect for intricate jazz lines.
7. Gibson Tal Farlow
• Overview: This signature model, released in 1962, honors jazz guitarist Tal Farlow. It’s based on the L-5CES with some unique modifications.
• Notable Players: Tal Farlow himself.
• Why It’s Great: Its distinctive design and tonal versatility make it a jazz classic.
8. Gibson ES-150
• Overview: Released in 1936, the ES-150 is often considered the first commercially successful electric guitar, paving the way for electric jazz.
• Notable Players: Charlie Christian.
• Why It’s Great: Its pioneering design and “Charlie Christian” pickup produced a warm, fat jazz tone.
9. Gibson Johnny Smith
• Overview: Introduced in 1961, this model was co-designed with jazz guitarist Johnny Smith, featuring a floating mini-humbucker for a pure acoustic tone.
• Notable Players: Johnny Smith.
• Why It’s Great: Exceptional acoustic-like sound with minimal feedback.
10. Gibson Howard Roberts Fusion
• Overview: A semi-hollow guitar with a fusion of jazz and rock elements, designed in collaboration with Howard Roberts.
• Notable Players: Howard Roberts.
• Why It’s Great: Its design is perfect for jazz fusion and contemporary jazz.
Honorable Mentions:
• Gibson ES-125: A more affordable archtop with a warm, vintage tone.
• Gibson ES-330: Similar to the ES-335 but fully hollow, offering a more acoustic-like tone.
• Gibson L-7: A step down from the L-5 but still a fantastic jazz guitar.
These models have defined the sound of jazz guitar for generations, making Gibson one of the most revered names in the genre.
-
Will A.I. argue with me like other forum members? Where’s the fun in that,Lol!
-
Heh heh...
Q. What are the limitations of using AI chat as a source?
Using AI chat as a source comes with several limitations, including:
1. **Accuracy and Reliability**: AI models may generate inaccurate or misleading information. They can sometimes provide outdated or incorrect facts, especially if the training data is not up-to-date.
2. **Lack of Expertise**: AI lacks specialized knowledge or understanding that a human expert might possess. It may not provide nuanced insights or deep contextual understanding on complex topics.
3. **Context Limitation**: AI models have a limited capacity to understand context fully. They may misinterpret questions or provide vague answers if the input is not clear or specific.
4. **Bias**: AI systems can reflect biases present in the training data, leading to answers that may reinforce stereotypes or provide unbalanced viewpoints.
5. **Absence of Critical Thinking**: AI does not possess reasoning or critical thinking abilities. Its outputs are based on patterns in data rather than comprehensive analysis or judgment.
6. **Temporal Relevance**: The information provided may be outdated, especially if the model's training data is not frequently updated. This is crucial for fields that evolve rapidly, such as technology and medicine.
7. **Lack of Source Attribution**: AI responses may not adequately cite sources or provide references for the information, making it hard to verify the reliability of the facts presented.
8. **Ethical Considerations**: Relying solely on AI for sensitive topics can lead to ethical issues, as the model may not handle such topics appropriately or sensitively.
9. **Generality**: AI tends to provide generalized responses that may not cater to specific needs or individual cases, leading to potentially irrelevant or unhelpful answers.
10. **Limited Interaction**: Unlike human experts, AI cannot ask clarifying questions to better understand user intent or provide personalized guidance.
11. **Dependency**: Over-reliance on AI for information can diminish critical research skills and the ability to evaluate sources independently.
For these reasons, it is advisable to use AI chat as a supplementary tool rather than a primary source of information. Always cross-check facts and seek expert opinions when necessary.
-
who was the guy that precicted the collapse of the internet for 1995? Originally Posted by Christian Miller
LLMs help me to do things that were totally out of my reach a year ago. i create websites and applications with php, java, html, python, etc., all without being able to write a single line of code. it's unreal. and we only get the breadcrumbs of AI, LLMs thrown out for the peasants.
the costs of running the *real* models are getting insane. the announced O3 model needs up to 10.000$ for a *single* task in high performance mode. and yet no bezos, zuck or musk, let alone the leading countries, can afford not to use these super AIs that are starting to show up.
AI will eventually take over most of banking, administration, call-centers, programming, and huge parts of production and logistics. the oligarchs will control all aspects of the productive force, including labor. good luck, old school national economies.
oh, and you can totally buy this today for 150k. or for 80k without wheels. no more schlepping that twin reverb.
-
Sure, and after Apollo people thought there’d be a moon base in a decade or two. It’s hard to make predictions. Science fiction writers and futurologists have always sucked at it, but at least the former realised they were writing fiction. Originally Posted by djg
who was the guy that precicted the collapse of the internet for 1995?
LLMs help me to do things that were totally out of my reach a year ago. i create websites and applications with php, java, html, python, etc., all without being able to write a single line of code. it's unreal. and we only get the breadcrumbs of AI, LLMs thrown out for the peasants.
the costs of running the *real* models are getting insane. the announced O3 model needs up to 10.000$ for a *single* task in high performance mode. and yet no bezos, zuck or musk, let alone the leading countries, can afford not to use these super AIs that are starting to show up.
AI will eventually take over most of banking, administration, call-centers, programming, and huge parts of production and logistics. the oligarchs will control all aspects of the productive force, including labor. good luck, old school national economies.
oh, and you can totally buy this today for 150k. or for 80k without wheels. no more schlepping that twin reverb.
Of course I might be eating these words in a year or two. We’ll see!
One reason Nuclear Power is suddenly a hot topic (across the aisle) is that Big Tech needs more energy for this stuff. It has been suggested that energy use is a bottleneck for AGI.
I don’t doubt that all of this will be disruptive to the world of work. It already has been. However the thing driving this is of course human beings and their relationship with and misunderstanding of the technology. The fact that it is not in fact fit for purpose does not mean it won’t be used. Just look at Tesla “self driving” and the Google AI feature. We are going to be saddled with a lot of AI slop and bad functionality. I mean that’s the reality now, right? Google and Facebook.
However having played with AI and my talked to my missus who uses it help with coding and so on it’s clear that it’s really useful, but incredibly flaky.
You can’t trust an LLM to program a computer, but it may make helpful suggestions among a certain level of slop. Again I doubt this will be rectified by increasingly sophisticated LLMs, because I really do think this is something that’s baked into the tech.
I might be totally wrong of course, but this is something I kind of feel on a deep level from using GPT, Midjourney etc. the latter is not a replacement for a competent human illustrator or graphic designer for example. Not because the imagery it produces isn’t very impressive but because ultimately it can’t give me reliably what I want beyond simple thumbnail design images, or elements I might be able to rework into something better. (Also everything it makes looks ‘midjouney-ish’ in composition and colour palette, although this may change, and may also be to do with my lack of skill in prompting it).
I have used it for imagery but I tend to use a lot less now than a year ago, partly because I see a lot of bad AI imagery out there and it looks a bit naff (which is not to say it can’t be used well, but that’s up to … guess what ? The human using it.)
Sora has similar issues as I understand it. It’s great at generating images for trailers where you only have to see the thing once for a few seconds.
(Of course, the money people probably won’t understand or care about this.)
People are looking at the trend of improvement of AI over the past few years and expecting it to continue like, say, CPU power did for decades. I certainly expect AI tools will become more sophisticated for sure, but very strongly suspect the issues they have are much more fundamental than you might hear from media. It always seems to be these holistic elements, narrative thread, musical form, artistic composition and so on that LLMs can’t do. As I understand it LLMs are also bad at maths.
So it’s important to disambiguate the actuality of what “AI” - which is impressive if limited - is how from the predictions of what it might do in a year or two - which is science fiction.
Which is not limited to AI, but tech generally. The start-up lifecycle. Musk is an extreme example, but he’s arguably the best at it, or at least has been.
we can look at the predicted timeline for Tesla, Space X etc as of, say, 2016 and contrast it to where we actually are with that technology. (This is a feature not a bug btw, we are just mugs for swallowing this stuff.)
I find it interesting that we’ve made machines that kind of dream rather than think. Who predicted that?
Sent from my iPhone using TapatalkLast edited by Christian Miller; 12-28-2024 at 07:47 AM.
-
“Warm, rich tone”? “Projection and clarity”? These are astounding observations from a source that has no more of a clue what they mean or if they’re accurate than did most of the material on which it was trained. AI is still a glorified parrot that tracks how often it hears everything and regurgitates the most often heard terms and phrases as the most correct.
i suspect that sensory input for AI is not only possible but already in development. Computers can already analyze sights, sounds etc. When a computer can listen to a guitar’s tone and identify it, I’ll have less trouble with it as a source of such information.
The sound of a Strad has been broken down and analyzed with far more sophistication than FFT provides. But no one (man or machine) has yet created a perfect duplicate electronically or in the flesh. So, at least for musical instruments, AI is still Cliff’s Notes. It has a long way to go before it becomes a sentient creator.
-
TL;DR LLMs remind me a lot of CST
When the only tool you have is a hammer…
Sent from my iPhone using Tapatalk
-
As a Software Engineer, I know that computer programming has become more automated over years. The software is writing more and more of the code.
-
Right now, if writing any kind of paper, thesis or project, ChatGpt can turn the time required from weeks to days, or from a week to a weekend. It can do coding and material research very efficiently. It is far behind what all these AI companies asking for funding say though. And that is a good thing, cause not much good will come of it. It will be one of the major threats for democracy and freedom in the 21st century. Maybe even worse.
-
Why?
Who doesn't want to get side-tracked and receive answers you never thought you wanted? 
I recently got a French letter from an acquaintance who's still learning the language so I'm used to getting messages that either have telltale errors or else are just awkward enough that you suspect a translator.
This letter was so perfect and well composed that I actually replied to it before I got suspicious. Sure enough, when I asked ChatGPT to write me a (French) letter carrying a comparable message I obtained a similar enough result that reads well enough like what an adult with a decent vocabulary and education would write. The one I received was a bit more personal; I assume I could have obtained something more like that if I'd written my message in Dutch and asked to translate that.
It's kind of eeky though. I'm still giving this particular acquaintance the benefit of the doubt but it showed me how easy it has become to fall for let's say emotional scam attempts these days.
-
AI is a tool. It's efficacy depends on the user, not the tool. If the user's not good enough, don't blame the tool.
-
As someone who has worked in AI for > 10 years, I’m pretty confident we’re at the verge of a large paradigm transformation in how humans work. Calling out examples of a model doing poorly on a niche problem is not a substantive criticism. These algorithms are data hungry and how well they perform on different tasks depends on the representation and quantity of data. I think a more effective way of using AI in its current state is as an assistant. More often than not it can get you to 80% of the solution and with additional prompting you can frequently get to 99% with little manual intervention and often 100%.
For software engineering, we’re not using it to rearchitect systems or build new software from scratch. But if you want a bunch of unit tests, basic migrations, or repetitive tasks like data engineering done quickly, there’s no reason for a human to be doing it manually instead of using AI.
-
That’s a cliché, not a rule - and it’s not applicable here. Try planing a smooth surface with a ragged blade or drilling a perfect hole with a bent bit. Originally Posted by ragman1
AI still has too high an error rate to be a reliable source of information for many (if not most) purposes. In addition to that limitation, it’s inherently biased because the foundation of its knowledge base is inherently biased.
-
For once I actually agree with ragman. And it does apply here. Error rate is an attribute of the tool specific to use case. The user should decide when it is appropriate and how to validate and QA the output.
-
Are you saying you can use AI to reliably write software for these simpler tasks? Or does a human still have to look over the code afterwards to check for and fix bugs? And, if so, how much time has really been saved? Originally Posted by omphalopsychos
-
The AI is both writing the code and looking for bugs. Humans are doing executive review but this is after the AI has already written tests. The result is 80% - 90% of time saved on these types of tasks. Originally Posted by CliffR
Another use case is for software engineering writers block. Writing code is like writing anything and people often get stuck while evaluating alternatives and coming up with somewhere to start. Using AI to jump start the writing process is also super effective.
-
Interesting and, I must say, surprising. I wish I could remember the source, but I seem to remember reading that the burden of fixing AI-generated code proved greater than the time otherwise saved. Still, not my area of expertise. I did find this interesting survey of techniques to verify correctness here: Benchmarks and Metrics for Evaluations of Code Generation: A Critical Review Originally Posted by omphalopsychos
-
Back in the early 90's the CEO of American Express resigned his incredibly well paid and powerful position to become the head of their newly minted Internet division. When asked why, he replied that the Internet was going to change the world as much as the industrial revolution had, and he wanted to make sure that his Company was at the forefront of the new era. At that time, I was skeptical of his assertion. He was correct. Now there are many who claim the same for AI. This time, my skepticism is tamped way down.
I just hope that Kubrick's film 2001, A Space Odyssey does not turn out to be prophetic.
-
Originally Posted by CliffR
All I will say is that the quality of Generative AI is advancing so rapidly that anything you read 1.5 years ago is probably completely outdated now.
A couple months ago one of my teams did a hack week project where they essentially cloned a data scientist, with a natural language interface. I.e. you can just ask it a question about your customers, historical trends, analytics, etc. And the bot writes a SQL query for you and executes it, returning the summarized analysis. The team also validated it against the expert data scientist who confirmed the queries it wrote were correct.
-
I think Back To The Future 2 is far more likely to define our future. As I see it, the transition has already begun. Originally Posted by Stringswinger
-
Don't get me wrong - I find it fascinating and awesome what can be achieved by LLM, but writing a working program is a long long way from writing an SQL prompt.
I just asked the latest Gemini to write a program to solve the n-queens problem (place n queens on an n x n board so that no two are attacking each other). It did surprisingly well, but there are (at least) three critical bugs:
Write a program in C to solve the n-queens
#include <stdio.h>
#include <stdbool.h>
#define MAX_SIZE 20 // Define a maximum board size
int board[MAX_SIZE][MAX_SIZE];
int n;
int solutions = 0;
// Function to print the board
void printBoard() {
for (int i = 0; i < n; i++) {
for (int j = 0; j < n; j++) {
printf("%d ", board[i][j]);
}
printf("\n");
}
printf("\n");
solutions++;
}
// Function to check if a queen can be placed at board[row][col]
bool isSafe(int row, int col) {
// Check the column
for (int i = 0; i < row; i++) {
if (board[i][col] == 1) {
return false;
}
}
// Check upper left diagonal
for (int i = row, j = col; i >= 0 && j >= 0; i--, j--) {
if (board[i][j] == 1) {
return false;
}
}
// Check upper right diagonal
for (int i = row, j = col; i >= 0 && j < n; i--, j++) {
if (board[i][j] == 1) {
return false;
}
}
return true;
}
// Recursive function to solve the N-Queens problem
bool solveNQueens(int row) {
if (row == n) {
printBoard(); // All queens are placed
return true; // Found a solution (important for backtracking)
}
for (int col = 0; col < n; col++) {
if (isSafe(row, col)) {
board[row][col] = 1; // Place the queen
if (solveNQueens(row + 1)) {
// If the recursive call returns true (a solution is found)
// we don't return here so that we can find all solutions
// Instead, we continue the loop to try other columns in the current row
}
board[row][col] = 0; // Backtrack: remove the queen
}
}
return false; // No solution found from this row
}
int main() {
printf("Enter the number of queens (n): ");
if (scanf("%d", &n) != 1 || n <= 0 || n > MAX_SIZE) {
printf("Invalid input. Please enter a positive integer less than or equal to %d.\n", MAX_SIZE);
return 1;
}
// Initialize the board
for (int i = 0; i < n; i++) {
for (int j = 0; j < n; j++) {
board[i][j] = 0;
}
}
if (solveNQueens(0)) {
printf("Total solutions: %d\n", solutions);
} else {
printf("No solution exists for n = %d.\n", n);
}
return 0;
}
-
The SQL query was an additional example. I don’t think writing homework problems in C is any harder but ok.
In your C program, now ask it to check for bugs and write tests. Then ask yourself, how long would it take for a CS college graduate to write the same level of code with testing?
Btw it doesn’t stop there, we’ve also got AI agents submitting pull requests and addressing build failures. I don’t know if we’re at the point where we can auto deploy to staging environments yet or run AB tests but I don’t see why not.
-
C is a much richer language than SQL, and therefore capable of hiding much more subtle errors - bad memory accesses, infinite loops, numerical issues, etc etc. BTW - I mis-spoke, I've only spotted two critical errors in the above segment. Rather than ask the AI to check for bugs (I wouldn't trust it) I think a more interesting question would be to ask: how long would it take a human programmer to detect and fix the bugs in the presented code? Did you spot them? Originally Posted by omphalopsychos
Reply With Quote
Recommandations for Hollowbodies for $600 and under?
Today, 05:20 AM in Guitar, Amps & Gizmos