I'm sorry to say, but in the both examples on the page I liked the original version more.
In the first example guitars sound very lively, like I'm actually listening to someone playing live. Very warm and cosy feeling. 2026 version is just some standard over-produced pop-music, that I would have completely ignored if I heard it "in the wild".
In the second example, 2003 vocal performance might be not perfect from a technique perspective, but I can hear real emotion in his voice. 2026 version sounds very bland, just like every other song that I might here on a pop-radio.
I want to say that I am not an anti-AI fanatic, I am generally curious about new ways to use technology (including gen-AI, but not limited to it) to create art. So I am not coming from "it's AI, therefore it's bad" perspective. I genuinely tried to listen with open mind and hear the music without thinking too much about how exactly it was created.
Basically, day (from sunrise to sunset) and night (sunset to sunrise) are each divided into 12 equal periods. But night hours and day hours are, of course, not equal to each other and changing throughout the year.
You can make a computer implementation by pulling astronomical sunset/sunrise times for a specific geographical location and then it's a simple arithmetic to convert from modern hours to horae temporales
Sure, there is some overlap. But there are also certain traits that would mean instant elimination from the gene pool in the wild but don't matter much when a human takes care of the horse.
I'm sure that if you investigate feral horse populations like North American mustangs or Australian brambies, you will find out that they don't "get ill if they breathe wrong". Because all that did, died without leaving any offspring long time ago.
> LLMs have seen humans act like conscious beings all over their training data because humans acting like conscious beings IS their training data.
How do we know that humans don't learn how to act conscious by observing other humans who act conscious?
Consciousness doesn't have a precise definition, but if you ask someone to describe it, there is a good chance that the description will include the concept of internal monologue.
The problem is that "internal" monologue is completely meaningless if you never heard an external monologue.
Also, people usually describe internal monologue as something that uses language and language is impossible to learn without communicating with other humans or at least observing other humans.
What I'm saying is that "well, LLM just pretends to be conscious, because it observed humans acting like conscious beings" doesn't really helps us to create a meaningful distinction between human consciousness and machine "consciousness", because same can be argued about us.
We don't know if feral children [0] are conscious and we don't know how to check it.
The idea that they really send spam to validate an email address sounds to insane to be believable.
Is it possible that they are somehow leaking the address to actual spammers?
For example, they (or the hypothetical email validation SaaS) use an infected email validation library that ex-fills every email supplied to it, or something like this.
Role tags are not actual symbols "<system>", they are special tokens that do not correspond to any normal text. So you can't really inject a role tag, that is not the actual problem.
> I can distinguish my own thoughts from your speech without effort; they arrive through completely different channels with completely different sensory signatures. But for an LLM, everything arrives through the same channel as one long token soup. Its own thoughts sit next to your instructions, which sit next to the contents of a random webpage it just fetched.
I was thinking about the original encoder-decoder transformers, that did have separate channels for input and their own output.
Why can't we bring it back? For example, one channel for system prompt and another for everything else.
By the way, if you are interested in nonograms specifically, there is a great website nonograms.org that has tens of thousands nonograms (both B&W and color) and no ads.
Interestingly, the pixelization/noise effect is applied clientside, so if you open an image in a new tab, you can see the original. Originals look much better, in my opinion.
Maybe it's a firefox bug that was introduced somewhere after 140? idk