Why ask forum members, when you can ask AI?

[Doctor Jeff]

I asked ChatGPT the best guitars for jazz. Here's what I got:

When choosing the best guitar for jazz, it's essential to consider factors like tone, playability, and aesthetic preferences. Jazz guitarists typically gravitate toward guitars with warm, rich tones, which are often semi-hollow or hollow-body electric guitars. Here are some top recommendations:

1. Gibson ES-175

• Why it's great: A classic jazz guitar, the ES-175 has been a staple among jazz players for decades. Its full hollow-body design and dual humbuckers deliver a warm, smooth tone ideal for traditional jazz.
• Best for: Straight-ahead jazz, bebop, and swing.

2. Ibanez Artcore Series (e.g., AF75 or AS200)

• Why it's great: Ibanez offers high-quality hollow and semi-hollow guitars at various price points. The Artcore series delivers excellent jazz tones with a modern edge and is affordable for beginners and professionals alike.
• Best for: Versatility and budget-friendly jazz options.

3. Epiphone Joe Pass Emperor II

• Why it's great: Designed in collaboration with legendary jazz guitarist Joe Pass, this guitar is an affordable alternative to high-end Gibsons. It offers a full-bodied, warm jazz tone with excellent craftsmanship.
• Best for: Jazz enthusiasts seeking a classic tone on a budget.

4. Gibson ES-335

• Why it's great: A semi-hollow design with a slightly more versatile tonal range, the ES-335 is perfect for jazz as well as blues and rock. It's a favorite among modern jazz players.
• Best for: Players who want a versatile guitar for jazz and other genres.

5. Gretsch G2420 Streamliner

• Why it's great: Gretsch hollow-body guitars have a unique twangy warmth that suits jazz, swing, and gypsy jazz styles. The Streamliner is a budget-friendly choice with great build quality.
• Best for: Players seeking vintage aesthetics and tone.

6. D’Angelico Excel Series (e.g., EXL-1)

• Why it's great: Known for their stunning design and impeccable jazz tone, D’Angelico guitars are a top choice for professional jazz musicians. The Excel Series offers the classic archtop sound with a modern touch.
• Best for: Traditional jazz and players seeking premium craftsmanship.

7. Eastman AR503

• Why it's great: Eastman is known for making hand-crafted guitars that rival high-end brands at a fraction of the cost. The AR503 is an excellent archtop guitar with a warm, resonant tone.
• Best for: A high-quality jazz tone on a mid-range budget.

8. Heritage H-575

• Why it's great: Built by former Gibson employees in the same Kalamazoo factory where classic Gibsons were made, Heritage guitars deliver a rich, traditional jazz tone with top-notch craftsmanship.
• Best for: Jazz purists seeking boutique-quality instruments.

---

Key Considerations:

• Pickups: Humbuckers or P-90s are standard for jazz, providing a warm, clean tone.
• Body Type: Hollow-body or semi-hollow guitars resonate better for jazz tones.
• Neck Profile: A comfortable, slightly thicker neck profile is preferred for intricate chord voicings.
• Budget: High-end guitars like Gibson and D'Angelico excel in craftsmanship, but brands like Ibanez and Epiphone offer affordable options.

Let me know if you'd like help narrowing down your choice based on budget or style!

[Doctor Jeff]

[Stringswinger]

The fact that neither the Gibson L-5, nor the Gibson Super 400 was included informs me that ChatGPT still has some learning to do.

[Cunamara]

Most AI stuff reads at best like an 8th grader who skimmed the Internet the evening before the essay was due.

I particularly enjoyed the stories of lawyers submitting briefs to courts written by AI, complete with hallucinated case reviews and nonexistent precedents. AI is not ready for prime time and when it is, will probably end us.

[nevershouldhavesoldit]

Most AI stuff reads at best like an 8th grader who skimmed the Internet the evening before the essay was due.

Great characterization! But there’s an associated factor that limits both of them - there’s no attempt (and in most cases, ability) to discern the quality of the information being skimmed.

AI models are “trained” on sets of data sourced largely from the internet. Machine learning engines are “fed” batches of data collated, labeled, and organized by their data engineering groups. Most such datasets reflect web search results, and there’s almost certainly ten thousand times more web activity about generally available Eastman, Ibanez, D’A, Gretsch etc than there is about L5s, S400s, Treniers, Campellones, Ungers, etc.

The staff who create the training datasets rely on what they find, and the machine that’s “learning” only starts off knowing what it was told. Fact checking and background research would take forever and cost a fortune. Nobody cares that this leaves out a huge chunk of information with which a bot could better educate itself and generate more accurate and usable answers.

So they work on the premise that if there’s less interest in something, it’s not as important. Then again, that answer about jazz guitars sounds like what you’d get if you asked a better than average sales person at GC etc

[Doctor Jeff]

The fact that neither the Gibson L-5, nor the Gibson Super 400 was included informs me that ChatGPT still has some learning to do.

If you even know what those are, I doubt you would need to ask AI about guitars...

Anyway, ChatGPT is for the 99% who aren't super-knowledgeable, not the 1% (or less) who are.

It *might* though bring up a brand you haven't thought of before though.

[Doctor Jeff]

Actually I got on a computer at the hospital today, and someone before me was using ChatGPT to compose discharge notes. This would just involve organizing information already available in the chart into a readable format.

Would require proofreading, of course, but I can see this as an acceptable use.

[Gitterbug]

How exactly did OP phrase the question? All-time best or currently available? I think AI gives a good summary of what's within reach today. While all Gibson archtops are currently extinct, it makes sense to highlight ES-175 for its seminal role and abundant second hand availability. The stock market looks at a share's free float. A handful of L-5s and Super 400s may change hands per year, genuine D'As or Strombergs even less so. To use a Finnish phrase, let's not pull a pea into the nostril at realizing AI doesn't use this Forum as its only source.

[Doctor Jeff]

How exactly did OP phrase the question? All-time best or currently available? I think AI gives a good summary of what's within reach today. While all Gibson archtops are currently extinct, it makes sense to highlight ES-175 for its seminal role and abundant second hand availability. The stock market looks at a share's free float. A handful of L-5s and Super 400s may change hands per year, genuine D'As or Strombergs even less so. To use a Finnish phrase, let's not pull a pea into the nostril at realizing AI doesn't use this Forum as its only source.

I just asked, "what would be a good guitar for jazz?"

As a bit of an aside regarding your Finnish phrase, I had a buddy in college who would put green peas in his mouth and shoot them out his nose. Ah the fun things we did in our youth...

[Christian Miller]

Basically Large Language Models are at their best when they have … errr… large amounts of language to train their models on. This is why they have a good general surface level knowledge and bad specialist knowledge.

(This btw is baked into this specific technology. It’s not a matter of refining the technology. LLMs are not AI and they are not a pathway to AGI - rather, they are interesting and sometimes useful tools with inherent limitations. Anything more is hype designed to keep Tech company share prices high.)

So asking an LLM standard gear questions involving well known brands will return sensible and accurate information, likewise it will tend to recommend mainstream and well known gear. Guitarists discuss that stuff all the time.

Asking an LLM about specialist archtop luthiers that sell a dozen guitars a year will return hallucinations because at this point it becomes a small language model.

Sent from my iPhone using Tapatalk

[pamosmusic]

The fact that neither the Gibson L-5, nor the Gibson Super 400 was included informs me that ChatGPT still has some learning to do.

Or that Chat GPT has a better handle on the average jazz guitarist's budget.

[nevershouldhavesoldit]

To use a Finnish phrase, let's not pull a pea into the nostril

Now we know why Finland is voted the world's happiest country every year. "There ain't no peas in our noses !! "

Speaking of AI, I Googled this to see if there was any info on the origin of the phrase, and there is not. But the first hit was AI's uninvited attempt to show how much it knew, to wit:

"A 'pea in the nose' means a small pea-sized object, like a pea itself, is stuck inside someone's nostril, considered a 'foreign body' in the nose, and usually requires medical attention to remove safely; if you suspect someone has something stuck in their nose, try gently blowing it out while blocking the other nostril, but if unsuccessful, seek medical help immediately; do not attempt to remove it with fingers or cotton swabs as this could push the object further in."

The moral of the story appears to be this: trying to remove a pea from your nostril is a fight to the Finnish.

[Spook410]

If you ask ChatGPT a shallow question, you'll get a shallow answer which is what we see here. Marketing copy carries a lot of weight in the response because it's easy to get. Input from places where the actual experts reside like this forum, are harder to tap into. ChatGPT doesn't know the difference and in this case, gave poor answers.

Around here when we encounter 'what's a good jazz guitar', the response is a flurry of questions. What kind of jazz? How much do you want to spend? What jazz players have a tone you like? What kind of amplification? et al.. Language models aren't there yet. And when they are, we may see responses tuned to the requirements of the corporations that control them.

[JazzPadd]

Interesting, thanks for posting that. It’s a little stuffy in terms of style, more like formal writing than casual forum speak.

i wonder how it’d respond to snarky or argumentative replies to its original suggestions?

I’ve yet to try AI, but I can see how it could be useful for tedious or superfluous formal writing tasks.

It seems to be big for student papers and essays. My nephew, who’s not a native speaker of English, uses AI for that. But not the free ones; he uses a paid one that can introduce grammatical and spelling errors to make the writing appear more student like.

[Christian Miller]

If you ask ChatGPT a shallow question, you'll get a shallow answer which is what we see here. Marketing copy carries a lot of weight in the response because it's easy to get. Input from places where the actual experts reside like this forum, are harder to tap into. ChatGPT doesn't know the difference and in this case, gave poor answers.

Around here when we encounter 'what's a good jazz guitar', the response is a flurry of questions. What kind of jazz? How much do you want to spend? What jazz players have a tone you like? What kind of amplification? et al.. Language models aren't there yet. And when they are, we may see responses tuned to the requirements of the corporations that control them.

(From what I have understood) LMM’s won’t be there ever. There’ll need to be a new technological paradigm of AI to make this possible.

As one knowledgable person put it - the problem with human interactions with LMMs is that we tend to associate linguistic fluency with intelligence. After all it’s our ability to talk that marks us out from the other animals. LMMs absolutely have this. Because of this we tend to infer a cognitive process or a sense of understanding of a subject as if it were a person. And tbf that’s the notion the Turing test was based on.

But an LMM is not a person or a mind in any sense. There is no cognitive process in the code of Chat GPT.

Except in the sense that it’s actually us.

The linguistic fluency of the LMM it is based on human linguistic fluency and what is said by millions of people. It has no inherent understanding of anything only which word is likely to follow another based on the prompt and the model. It’s a fancy mobile phone auto complete. It is at best a distillation of a vast swathe of real human communication (which includes of course both ignorance and knowledge.)

It cannot ever understand the query at all, let alone in that depth. It can only give a likely answer based on the language it has sampled from millions of examples of human generated text. More examples might make it better, but these things have already been trained on vast data sets.

Philosophically it’s INCREDIBLY interesting. Technologically it’s a development of systems that have been used for quite a while.

Sent from my iPhone using Tapatalk

[Gitfiddler]

My guess is that AI-driven Chat GPT obtained some of its LLM info from this forum. Yes, we humans are the true source of all of this wonderful technology that we call 'artificial intelligence'. AI does not create anything. It gathers and assembles our data and organizes it for our intended purpose.

That many be why the Chat GPT response reads like an extremely well-versed reader of the Jazz Guitar Forum. :-)

[Christian Miller]

My guess is that AI-driven Chat GPT obtained some of its LLM info from this forum. Yes, we humans are the true source of all of this wonderful technology that we call 'artificial intelligence'. AI does not create anything. It gathers and assembles our data and organizes it for our intended purpose.

That many be why the Chat GPT response reads like an extremely well-versed reader of the Jazz Guitar Forum. :-)

You know what they say - GIGO haha

No one comes here haha. That’s why I like it… it’s all humans.


Sent from my iPhone using Tapatalk

[Flat]

AI models are “trained” on sets of data sourced largely from the internet. Machine learning engines are “fed” batches of data collated, labeled, and organized by their data engineering groups.

And in the future, when the source data is itself nothing but data that has been generated by previous LLMs, the ouroboros of our species' demise will be birthed. Except this ouroboros will have a twist, and instead of being a symbol of infinity, it will become a symbol of nothingness. Let's hope the engineers can figure out how to program a sense of humor into these beasts.

[rpjazzguitar]

In terms of identifying specific brands, it did a better job than I could. It did seem to confuse vintage D'Angelicos with the later guitars that appeared under that name.

Every time I see a response like this I'm in awe that Chatgpt could do as much as it did. I'm well aware that it often gets things wrong, including simple facts. But, it does usually seem to understand the question, which is astonishing in and of itself. And, the technology is going to continue to develop.

For us geezers who programmed in languages like Fortran, Cobol, PL-1 and Basic, it's really hard to understand how to approach getting a computer to do this. I'm aware that it's a neural net approach, not IF-THEN statements. And I even have a primitive understanding of what that means. But, it's still amazing.

[vernon]

To answer the OP question…Because AI doesn’t play guitar.

[WilliamScott]

I asked ChatGPT
what benefit could come from putting peas in one's nose?

ChatGPT replied:


Putting peas in one's nose is generally not advisable, but if we consider it from a purely hypothetical or humorous perspective, here are a few quirky "benefits" that might be imagined:

1. Experimentation: A child or curious person might learn about cause and effect, discovering why it's a bad idea to put things in their nose.
2. Attention-Seeking: It might serve as a quick (albeit questionable) way to get attention from nearby adults or friends.
3. Unique Icebreaker: If shared later as a funny story, it could spark laughter and bonding in social situations.
4. Medical Education Opportunity: It provides a learning experience for medical professionals or parents on safely removing foreign objects from noses.

That said, peas in the nose can lead to choking, infections, or nasal blockage. So it’s best to avoid it unless you’re a small child who didn’t hear this advice yet!

ChatGPT

[Litterick]

Imagine somebody on this forum saying, "A semi-hollow design with a slightly more versatile tonal range, the ES-335 is perfect for jazz as well as blues and rock." Imagine somebody saying that at a dinner party. We would think them at best a stooge for the company, at worst someone incapable of independent thought. "Did you read that on Amazon?", we would ask. We expect more from our peers than ad copy.

[icr]

WHY ASK FORUM MEMBERS, WHEN YOU CAN ASK AI?

Isn't the AI bot trained on internet forums? You posted it so its YOUR post. You are now responsible for the content and its source. With your post, you have essentialy lobotomized yourself, cutting off your brains higher functions; throwing away millions of years of evolution in faovor of the AI bot.

[nevershouldhavesoldit]

Imagine somebody on this forum saying, "A semi-hollow design with a slightly more versatile tonal range, the ES-335 is perfect for jazz as well as blues and rock." Imagine somebody saying that at a dinner party. We would think them at best a stooge for the company, at worst someone incapable of independent thought. "Did you read that on Amazon?", we would ask. We expect more from our peers than ad copy.

Language evolves through use

[djg]

(From what I have understood) LMM’s won’t be there ever. There’ll need to be a new technological paradigm of AI to make this possible.

didnt hopfield just win a nobel prize for his work on emergent properties of simple systems, aka "the sum is larger than its parts"?