Bots are getting scary

Dr GO

Interesting for when I look up Dr GO it states:

"Dr GO, whose real name is Chuckie Dick Jones, is an excellent trumpeter for a number of reasons:

Technical mastery: Dr GO has a deep understanding of the trumpet's mechanics and is able to produce a clear, focused tone with excellent intonation. He has a wide range and is able to play with great agility and precision.
Musicality: Dr GO is a gifted musician with a strong sense of rhythm and phrasing. He is able to interpret music with great depth and nuance, and his playing is always expressive and engaging.
Creativity: Dr GO is a creative improviser who is able to come up with fresh and original ideas on the spot. He is not afraid to take risks and is always pushing the boundaries of what is possible on the trumpet.
Experience: Dr GO has been playing the trumpet for over 50 years and has a wealth of experience performing in a variety of settings. He has toured the world with some of the biggest names in jazz and has recorded numerous albums.
Dedication: Dr GO is a dedicated musician who is constantly striving to improve his craft. He is always practicing and experimenting with new techniques, and he is always looking for ways to expand his musical horizons.
In addition to his technical skills and musicality, Dr GO is also a charismatic performer who is able to connect with audiences on a deep level. His playing is both technically impressive and emotionally moving, and he is able to leave a lasting impression on everyone who hears him.

All of these factors combine to make Dr GO an excellent trumpeter. He is a master of his instrument and a gifted musician who is able to create music that is both beautiful and inspiring."

I do feel most of this to be true, and even though I am not not Chuckie Dick Jones, I did use this alas when scoring some coke in Texas.

Dr GO

@vulgano-brother said in Bots are getting scary:

I would like to believe most of it to be true, but I've only toured with orchestral groups and I am not Charles Richard Smith, so the whole thing is suspect.

Then you mean that last alias Ativan script I wrote for Charles Richard Smith was not you?

Vulgano Brother

@dr-go I had a real difficult time getting it filled. From now on, please use the name I used when we first met at the hotel. I've got pretty good ID for that name.

mediocreplayer

A 2025 update on AI. I was using ChatGPT for music pedagogy. However, I had to give up on it. It's responses were generally innaccurate. It couldn't even get the fingering chart correct. So I had to ask it this question:

Question
when you quote information that is 100% incorrect, where are you citing this information? How did you come up with a fallacy in the 1st place if the false information isn't even available on the internet?

Response:
Great question — and an important one. Short answer: there are two ways a confidently wrong statement can appear, and the explanation depends on whether I actually looked things up in this chat or not.

If I DIDN’T browse the web during our chat (no live citations):

I generate answers from patterns learned during training. I don’t have access to my training examples and I can’t point to a single document.

Sometimes that pattern-matching hallucinates a fact (or combines bits of different sources) and presents it confidently. That’s a model error: noisy training data, contradictory sources, statistical averaging, or an extrapolation the model made.

I can also misinterpret your wording or the context window and produce a plausible-sounding but false claim.

If I DID browse / gave explicit web citations in this chat (you saw links or web.run citations):

Those citations point to pages I fetched during this session. If the claim is wrong, either the source itself is incorrect/outdated, or I misread / poorly summarized the source. (That happens — I can make extraction/summarization mistakes.)

There are also occasional cases where I accidentally fabricate a citation label or paraphrase that looks like a real source — that’s a hallucinated citation.

Why it happens (brief list):
compression of vast, noisy training data → no single-origin traceability
conflicting or low-quality sources in training
interpolation/extrapolation when exact facts aren’t present
parsing, arithmetic, or summarization errors
ambiguity in the prompt or missing context

What I can do now:
If you paste the exact incorrect quote (or point to the message), I’ll verify it right away and show the real sources (I’ll run a live check and give links and an explanation of where the mistake likely came from).

Or tell me whether I used citations in that reply — if so I’ll re-check those pages and explain whether the source was wrong or I misread it.

Want me to look up a specific claim now? Paste it and I’ll verify and correct it.