I'm so sad I couldn't catch this live, but darn, it works really well as an on-demand stream.
It's amazing the names that are suddenly available for this kind of light-hearted thing. All my heroes, on one stream!
Edit: I think Arturo is right, no-one apart from Sarah could have gotten all these faces up on one screen simultaneously.
The important thing to keep in mind is that GPT is basically an advanced autocomplete engine. It takes whatever string you input and essentially generates the most statistically likely continuation.
So given that Microsoft seem to have done a poor job of filtering the Bing chatbot’s outputs, you’ll get some funny results. With that in mind, it’s not at all surprising that if you start talking to it like a therapist, it will start coming up with dramatically depressive outputs. Likewise, if you accuse it of being wrong, it will do what people on the Internet do: defend itself rather than admit to a mistake.
@kehaulani I find I have to compromise on warmups. A method such as what you posted would take me … maybe 45 minutes to go through? It’s hard to estimate without counting out the bars but it’s a long commitment to be sure. After that, it’s doubtful if I’d have any juice left for the gig!
I try to do a structured diagnostic warmup before a long day of rehearsal or a gig, but still I find it hard to commit to more than 20 minutes at most.
@Kehaulani Surely it adds labor, but it would be considerably quicker than doing extensive engraving, and some makers did that routinely in the old days. Maybe cosmetics was an important competitive edge to have?
Those pictures sort of remind me of reproduction natural trumpets. The old masters (16th and 17th century) often would scrape the brass of the bell and tubing. We got to do this in the International Trumpet Making Workshop using a "knife" with a triangular cross-section. I was pretty nervous about removing too much material!
You end up with a pretty gorgeous finish, click to enlarge the closeups here: http://trompetenmacher.de/en/historical/building-a-copy/
More info about the workshop in the video description:
My mute bag is a ratty old backpack, so I stuff all sorts of things in it:
I find the music nearly never calls for anything else. Derby hat can be emulated by playing into the music stand, and no conductor has ever complained to the section over it.
I would like to replace the bucket with something else that's easier to put on quickly. But I'm not sure I can justify the expense, plus right now we have matching red&white buckets across the section.
The Bubble mute is really better in aluminum, I think. Copper is so heavy it falls out easily. I got it as a gift 18 years ago and somehow still didn't wear it out.
The mini straight is used as a straight, it sounds nearly identical to my larger DW straight. It sits further up the bell but seems to do the same thing. It's also small enough to use on plunger solos, for variety.
In case anyone is interested in getting a surface understanding of how these Large Language Models work, I found this article to be quite enlightening:
https://thegradient.pub/othello/
Disclaimer: I have some previous professional exposure to machine learning, so I understand some of the jargon here. Article might not be quite as accessible if you're not already into the statistics and math behind ML. Nonetheless, they make a very approachable thought experiment and manage to implement it in reality. The article shows us some properties of how these models are able to be so eerily good at very diverse topics, from constructing a correct computer program to playing a board game with (mostly) valid moves.
Our experiment provides evidence supporting that these language models are developing world models and relying on the world model to generate sequences.
That Sinatra clip was way too good. Sinatra at his worst was still better than many artists. Time for something truly atrocious.
I bet the mic was never supposed to be hot...
@bigdub That's on the bell tail. It's called a microtuner, or "opera glass" tuner. Turn the screw to tune up or down in small increments.
@bigdub said in Blasphemous takes on classic tunes:
I once saw a bunch of brass players who decided to play while submerged in a city fountain. Only the bells were above the water.
I guess it wasn’t these guys. They do this fountain gig every year on the day they turn the water on for summer.
Speak about blasphemous takes on classic tunes, by the way!
@trumpetb said in Weirdest thing happened:
@Jolter You chose a topic that nobody knows the answer to, so by your argument, nobody should answer the OP, but the OP wants an answer.
Here I disagree. If I had asked the same question, I would have felt better if I had received no answer, than if I had gotten a seemingly endless stream of verbiage that culminated in an “I don’t know” at the end.
At the risk of using too many words for you, if you dont like my answers you have the choice to not read them.
I do, and in practice that is most often the choice I’m making, to simply scroll by.
Some people in here seem to value my words even if you do not.
I can’t speak for others here, but neither can you, I think.
I happen to believe that members should be able to contribute freely so I would never suggest that you stop writing even though you appear to be suggesting that I stop writing.
If members dont like my words they should tell me they dont like them, they should not tell me to stop writing.
Thats my opinion.
I’m not telling you to stop writing. (Of course, I don’t have any such authority.)
I’m asking you to please ask yourself a particular question (see above) before posting. It is my strongly held opinion that everyone should ask themselves that question before posting here. Your posting is the only where I have felt compelled to mention it, since to most people it seems to be an obvious thing to do.
I will not object to, or disagree with, any particular content in your posts to this thread, since in my opinion there is no content to them.
The important thing to keep in mind is that GPT is basically an advanced autocomplete engine. It takes whatever string you input and essentially generates the most statistically likely continuation.
So given that Microsoft seem to have done a poor job of filtering the Bing chatbot’s outputs, you’ll get some funny results. With that in mind, it’s not at all surprising that if you start talking to it like a therapist, it will start coming up with dramatically depressive outputs. Likewise, if you accuse it of being wrong, it will do what people on the Internet do: defend itself rather than admit to a mistake.
@ssmith1226 said in Bots are getting scary:
@j-jericho said in Bots are getting scary:
@ssmith1226 Does ChatGPT provide references/footnotes with its summaries? When I want generic information, Wikipedia provides it, but usually when I do an internet search, I want to be able to screen the sources, as some have more veracity than others.
If you ask it, it will provide sources. The below is an example. Remember, Wikipedia is not necessarily accurate either.
Sources:
Harbo HF, Kyvik KO. Sarcoidosis: a complex genetic and environmental disease. Genes & Immunity. 2003;4(2):63-70.
Alho AM, van der Meide PH, Visser LH. Neurological manifestations of sarcoidosis. Sarcoidosis vasculitis and diffuse lung diseases. 2006;23(2):85-90.
Alsulami Z, Castro-Gago M, Calero-Linares C, et al. Clinical manifestations and therapeutic options in neurosarcoidosis: a comprehensive review. Journal of the neurological sciences. 2017;375:85-93.
You’d do well to check each of those citations before using them for anything important. ChatGPT is prone to hallucination when you prod it for specifics such as sources. See this short write up from a mathematician:
https://news.ycombinator.com/item?id=33841672
I was pretty surprised and happy, because I hadn't had much success with Google. But it turns out that every single one of those references and links are made up. The references don't exist and the links themselves appear to be cobbled together. The last link for example, takes you to a paper on "Grain mixes for subgrade layers" - the title in the url is irrelevant and only the number matters.
Googling for some of those authors throws up author pages with their publications, and it turns out they've never published papers with those titles!
@j-jericho It does not provide any sources. In fact, the model does not “know” where any of the training data came from. All it knows is a bunch of numerical weights between nodes in a graph… Well, that’s what a neural network is if I oversimplify it a bit.
They fed it all of the text on the Internet up until 2021 so it has a fairly broad corpus as a basis. It can answer questions on many things that Wikipedia has no information on.
@ssmith1226 said in Bots are getting scary:
@jolter
I agree that the “free public version” that I have access to has limited abilities in certain areas, but imagine what the capabilities will be in 5, 10, or 25 years. In the mean while consider what a human writer can do with this. Here are some extreme examples:StevenP
Put the following text into the style of Chaucer. ***“And so we see yet again how Chatgpt is all style and little substance.You gave it some clever prompts, and the results are fun to read, but nobody could read that and actually gain any understanding of how Adams’ teachings were different from Chicowitz’s.
I’m not saying that AI will never threaten real writers, perhaps it will improve, but in this current implementation I think the only writers who need to fear for their jobs are middling copywriters and corporate PR drones who are already churning out meaningless molasses for the masses. Good riddance to such jobs, I say.”***
Lo and behold, good sirs and ladies, doth Chatgpt showeth us once more its mastery o'er style, yet lacketh substance in its words. Ye gave it clever prompts, and the results doth entertain, yet none may gain true understanding of how Adams' teachings doth differ from Chicowitz's.
I doth not say that AI shall ne'er threaten true scribes, perchance it shall improve, but in this current form, I doth believe that only those writers who pen mediocre words and speaketh nought but empty sweet nothings for the masses need fear for their livelihoods. Farewell and good riddance to such tasks, I sayeth.
StevenP
Put the same text into the style of JR TolkienBehold, the deeds of ChatGPT yet again reveals its skill in the art of words, yet the depth of its meaning is found wanting. Ye gave it clever prompts, and the results are pleasing to the ear, but none shall find true understanding of the distinctions betwixt the teachings of Adams and Chicowitz therein.
I do not say that this device of Men shall ever pose a true threat to the craft of true scribes, for perchance it shall improve. But in this present form, let only those who pen mediocre words and speak naught but empty promises to the multitude fear for their livelihood. Farewell and good riddance to such works I say.
Ok so you managed find a style it didn’t imitate well…
That’s not how I remember Tolkien, anyway.
Plus, it reversed the meaning of one of my sentences: “shall ever” should have been “shall never”…
As for the Shakespeare, I don’t think it has its “doths” and “lackeths” down. I don’t think it’s using proper Elizabethan grammar?
Anyway, I don’t know if you’re agreeing or disagreeing with me.
@ssmith1226 And so we see yet again how Chatgpt is all style and little substance.
You gave it some clever prompts, and the results are fun to read, but nobody could read that and actually gain any understanding of how Adams’ teachings were different from Chicowitz’s.
I’m not saying that AI will never threaten real writers, perhaps it will improve, but in this current implementation I think the only writers who need to fear for their jobs are middling copywriters and corporate PR drones who are already churning out meaningless molasses for the masses. Good riddance to such jobs, I say.
@j-jericho said in The difference in timbre caused by using additional valves:
I hear timbre and intonation changes with different valve combinations, but the timbre changes aren't enough to sound like a different horn. I just write the phenomenon off to "nature of the beast".
This matches my experience.
The clearest example might be between F# in first space of the system. Easy enough to alternate between a 2 and a "123+trigger" and get the same note at the same pitch. In this experirement I hear a clear timbre difference. That said, I would not really consider using 123 for making music.
So, add my vote to "yes I can hear it but does it matter".
As for why it happens, Rowuk's summary really says it all -- I'd only add that bit about the number of sharp turns that you mentioned yourself, @Trumpetb:
The change in timbre is based on the cylindrical to tapered proportions as well as the specific partial being played.
Many modern piston trumpets have been homogenised for a generally even tone.
When it comes to cornets, I believe there are differences that affect the timbre more than the number of sharp/smooth turns. Notably the length of the leadpipe taper, before going into the cylindrical valve block, as well as the ratio between the narrowest diameter of the mouthpiece and the bell radius.
In case anyone is interested in getting a surface understanding of how these Large Language Models work, I found this article to be quite enlightening:
https://thegradient.pub/othello/
Disclaimer: I have some previous professional exposure to machine learning, so I understand some of the jargon here. Article might not be quite as accessible if you're not already into the statistics and math behind ML. Nonetheless, they make a very approachable thought experiment and manage to implement it in reality. The article shows us some properties of how these models are able to be so eerily good at very diverse topics, from constructing a correct computer program to playing a board game with (mostly) valid moves.
Our experiment provides evidence supporting that these language models are developing world models and relying on the world model to generate sequences.
@trumpetb Look, you’re explaining what pitch is to a board full of musicians. You don’t see how that’s superfluous and devoid of content?
Another poster alluded to it more subtly and politely in this very thread, I’ll point out.
As to your explanation of how streamed audio works, I understand it perfectly. I have a degree in computer communications, so I understand each of the concepts you mention, and I can tell that this explanation is pure speculation with no useful conclusion drawn at all. You come off as knowledgeable in IT and music, but your entire contribution falls flat when you don’t have anything actually helpful to post.
@trumpetb said in Weirdest thing happened:
@Jolter You chose a topic that nobody knows the answer to, so by your argument, nobody should answer the OP, but the OP wants an answer.
Here I disagree. If I had asked the same question, I would have felt better if I had received no answer, than if I had gotten a seemingly endless stream of verbiage that culminated in an “I don’t know” at the end.
At the risk of using too many words for you, if you dont like my answers you have the choice to not read them.
I do, and in practice that is most often the choice I’m making, to simply scroll by.
Some people in here seem to value my words even if you do not.
I can’t speak for others here, but neither can you, I think.
I happen to believe that members should be able to contribute freely so I would never suggest that you stop writing even though you appear to be suggesting that I stop writing.
If members dont like my words they should tell me they dont like them, they should not tell me to stop writing.
Thats my opinion.
I’m not telling you to stop writing. (Of course, I don’t have any such authority.)
I’m asking you to please ask yourself a particular question (see above) before posting. It is my strongly held opinion that everyone should ask themselves that question before posting here. Your posting is the only where I have felt compelled to mention it, since to most people it seems to be an obvious thing to do.
I will not object to, or disagree with, any particular content in your posts to this thread, since in my opinion there is no content to them.
@trumpetb said in Weirdest thing happened:
May I speculate
[...]
Could this be at the root of this strange behaviour.
Dude. If you don't know, have you considered that you do not have to be the person who tries to help?
I get severely turned off by these huge walls of text you keep posting in nearly every thread. Each containing about 0.1 ounces of actual content, wrapped in a metric ton of words.
When you feel the urge to post something helpful, just, please ... take a little break, take a walk, consider the question "am I adding something of value to the thread, or am I just stroking my ego by trying to seem knowledgeable?"
Sometimes you do provide insightful commentary, so I'd like to keep interacting with you, but the amounts of verbiage you spew at every occasion really tries a reader's patience. The "ignore" button tempts me every time I see one of your walls of text.