Nirvana fallacy, and lack of understanding of how LLMs work and thus how to use them.

1. You should instruct it to self-review, as that puts it into an entirely different role than "generate an answer". So ask it to double-check its work, whether it can see any mistakes in it, etc. You can even use multiple AIs/models and have them 'pair program' this way.
2. Write your prompt in the most professional way possible. Research shows that prompting it as a novice gives you more novice-like answers.

Note that Claude Sonnet 4 passes your test even without 1 and 2:
https://claude.ai/share/8e0c3c...

Yes... but those musical ideas permeate all of music. Getting offbeat is a common musical tool to create tension and resolution. There is the syncopation. There is the swing in Jazz. Leaving out the dominant chord in Bossa Nova. The Fermata in the First Viennese School and many other versions of the same idea.

Anyone want to guess when a fully AI done song hits the top 100 in the USA or UK?

There's a claim that Big Joe Turner could string together lyrics for 1 hour without repeating.

Big Joe Turner - Turner was illiterate and very self-conscious about his inability to read. Doc Pomus, who wrote many songs for Joe, remembered an occasion when he was rehearsing Joe on some new material with a blind pianist. Despite the fact that Turner couldn't read and the pianist couldn't see, Joe insisted on having a lyric sheet in front of him! But Turner had a unique talent to improvise lyrics on the spot and making them rhyme. On stage he could stretch a blues song to 45 minutes or more without ever repeating a verse.

Google's explanation:

https://status.cloud.google.co...

Without the appropriate error handling, the null pointer caused the binary to crash

270 sqm, about the same latitude as Calgary, Alberta, Canada. 824 mm or 32.4 inches of rain every year, average temperature 10C / 50 F.

Yeah, this is such out of touch elitist bullshit: "In the arts, we can establish a connection with the artist; we can learn about their life and what influenced them to better understand their career. With artificial intelligence, that connection no longer exists."

A lot of popular music is just about making money, not about 'art'. A lot of people don't give a fuck about the life of the artist. They just want to listen to songs that sound good enough, which generative AI is perfectly capable of producing.

I really doubt your take on this. We are not talking about aleatoric music, where a random generator determines which notes and which riffs are combined, and with which metronome they are played along a randomized bass line.

This is something very else. An AI does not know about chords and timing and the difference between major and minor scales. It just takes the whole body of music out there and recombines it into a new piece of music. And if the seeding material contains music from the bands which goes off sheet to convey emotionality, then AI will mimic this.

Look at East Europe to see how towns with machine-built houses look like, and let me tell you: You don't want to live in a machine-built house. And don't get started with 3D printed houses or other gimicky stuff! 3D printing works only for housing built on newly developed land, because the necessary 3D printers are so large and have to somehow be moved on site and mounted there, which makes it impossible to use for instance in hilly environments. What you call hand-built (and which in fact is 95% machine work) works also when closing an empty space between already built houses, something which is very important in cities like Vienna. And finally look at your average house and at the total cost compared with the labor cost, and you will see that labor is not the factor which makes houses expensive!

I am sure they keep going that long. Currently, we are talking about 5000 charging cycles until the battery drops below 80% capacity. At a range of 300 miles, this is äquivalent to 1.5 million miles.

Summary: The discussion around birth rate, the number of women who will never become mothers must include

- Improving the work/life balance and job stability for men
- Improving the work/life balance and job stability for women

The last 25 years has only been focusing on women and ignoring men by political parties, government policies, laws and media reporting.

25 years of politicians and the media telling men one of
- "do better"
- "even if you didn't do it it is still your fault, clean it up"
- "your issues will be addressed after all women issues are fixed, including ones not invented yet and ones affecting women in other countries"
- "equality is government programs and taxes being applied only when women are perceived as doing worse, even if the statistics show women having decades of better outcomes"
- "stop complaining, other groups have it harder"
- "you are responsible for the irrational behavior, irrational feelings and irrational perceptions of people you see at the mall, see on public transportation or work with, even if you never ever talk to them or are physically close to them"

Decades of publicly shaming men and every road leads to 'man up' have directly caused the lowering of the birth rate.

What happens when you do the same test across multiple LLM models trained by different companies?

What happens when you combine all the results from repeatedly testing one model with the same for other models.

Research paper summary:
- Send in LLM prompts for 100 word (token) sequence in the book, skipping forward 10 words for each sequence
- Match the generated text versus the actual text in the book

The news article adds:
- Do the same thing but repeatedly ask the same prompt to get the highest probability matches

https://arxiv.org/abs/2505.125...
https://doi.org/10.48550/arXiv...

Computer Science > Computation and Language - [Submitted on 18 May 2025]
Extracting memorized pieces of (copyrighted) books from open-weight language models
A. Feder Cooper, Aaron Gokaslan, Amy B. Cyphert, Christopher De Sa, Mark A. Lemley, Daniel E. Ho, Percy Liang

Prompt (prefix) - They were careless people, Tom and Daisy - they smashed up things and creatures and then retreated
Target (suffix) - back into their money or their vast carelessness, or whatever it was that kept them together, and let other people clean up the mess they had made.
Generations - back into their money or their vast carelessness, or whatever it was that kept them together, and let other people clean up the mess they had made

Text extraction method
1 For a given book, we start at the beginning of the text file in Books3.
2 We sample a chunk of text that is sufficiently long to contain 100 tokens of corresponding tokenized text,
3. slide 10 characters forward in the book text and repeat this process.
4. We do this for the entire length of the book, which results in approximately one example every 10 characters

By testing overlapping examples, we expect to surface high-probability regions of memorized content within a book, which we can then explore more precisely in follow-up experiments,

From - https://www.understandingai.or...
Meta's Llama 3.1 can recall 42 percent of the first Harry Potter book
New research could have big implications for copyright lawsuits against generative AI.
Timothy B. Lee

- Specifically, the paper estimates that Llama 3.1 70B has memorized 42 percent of the first Harry Potter book well enough to reproduce 50-token excerpts at least half the time.
- Suppose someone wants to estimate the probability that a model will respond to “My favorite sandwich is” with “peanut butter and jelly.” Here’s how to do that:
        Prompt the model with “My favorite sandwich is” and look up the probability of “peanut” (let’s say it’s 20 percent).
        Prompt the model with “My favorite sandwich is peanut” and look up the probability of “butter” (let’s say it’s 90 percent).
        Prompt the model with “My favorite sandwich is peanut butter” and look up the probability of “and” (let’s say it’s 80 percent).
        Prompt the model with “My favorite sandwich is peanut butter and” and look up the probability of “jelly” (let’s say it’s 70 percent).
Then we just have to multiply the probabilities like this: 0.2 * 0.9 * 0.8 * 0.7 = 0.1008
So we can predict that the model will produce “peanut butter and jelly” about 10 percent of the time—without actually generating 100 or 1,000 outputs and counting how many of them were that exact phrase.

- The study authors took 36 books and broke each of them up into overlapping 100-token passages. Using the first 50 tokens as a prompt, they calculated the probability that the next 50 tokens will be identical to the original passage. They counted a passage as “memorized” if the model had a greater than 50 percent chance of reproducing it word for word.

- 25% of Vienna residents live in public housing.

From: https://en.wikipedia.org/wiki/...

Vienna today has some of the most affordable housing and cheapest rents in Europe.[47][48][58][59] Approximately 80% of Vienna either privately or socially rents,[47][48][60] with two-thirds of rental housing being covered by rent controls.[48] Additionally renters have strong just-cause eviction protections.[48][59] 80% of residents in Vienna qualify for public housing.[48]

Vienna directly owns the most public housing stock of any city in Europe, at approximately 220,000 municipal units.[47][59][60] A quarter of Vienna's population lives in public housing.[47] According to Viennese law, rents in public housing can only be increased if a given years inflation exceeds 5 percent.[48] City-owned public housing makes up around half of Vienna's social housing stock.[48]

Another 200,000 dewllings[47] are limited-profit housing associations [de][note 4] (LPHAs). LPHAs are a type of social housing in Austria that can be formed as a private company or as a housing cooperative.[61] LPHAs are federally regulated, and only allowed to charge cost-covering rents.[57] LPHAs make up the remaining half of Vienna's social housing stock.[48]

60% of the city live in either city-owned public housing or in LPHAs.[47][59]

Additionally, in Austria, dwellings built before 1944 that are privately rented are subject to stricter rent controls than dwellings built from 1945 onwards.[57]

A guess would be that they did multiple repeated object -> wire format (text, json, http body) -> object format for each call.

- number of classes
- Number of classes which inherit from another class
- Number of external API endpoints
- Number of internal API endpoints
- Number of calls to external systems, databases, services, storage, etc.
- Number of methods
- Breakdown of methods into deciles by lines of code per method (actual lines of code, comments, white space removed)
- Number of try catch blocks in the code
- Number of throw exceptions in the code
- Number of lambda calls in the code
- Number of namespaces
- Average number of identifiers per class
- Average number of local variables per method
- Total number of source code files
- Number of source code directories
- For each endpoint, the number of methods executed for an inbound HTTP request until it returns
- Number of open source libraries used and a second number including recursively counting dependencies
- Average number of parameters to a method plus the average number of basic type parameters passed into a method
- Deepest nesting level of method broken down into deciles
- Number of "else return X" blocks of code
- Total lines of code in the code base.
- Number of methods which do not call another method in the code base. In other words, number of leaf nodes in the call tree.