And those corrections require atomic clocks.

You think LLMs would write "Slashdot doesn't have table support, so it's hard to do side-by-side well, but:"? You think LLMs like writing sentences that end in conjunctions that express incomplete thoughts? Do you think LLMs write "high frequences", "naively pairs", call H.265 X.265 (I was thinking of what you refer to it as in ffmpeg), imbalanced parens after "lacks motion vectors",etc?

Don't get me wrong, I am a LLM user, I absolutely do use it sometimes for search, fact checking things I'm writing, spelling/grammar checks (clearly not that time, lol), etc. But I write my own posts.

Slashdot doesn't have table support, so it's hard to do side-by-side well, but:

JPEG XL advantages:

Licensing:

JPEG XL: Fully royalty-free (Open source / Apache 2.0)
HEIC: Proprietary patent minefield (MPEG LA, Advance, Velos; royalties apply)

Legacy JPEG Migration:

JPEG XL: Lossless, reversible bitstream transcoding (~20% smaller); fast enough for real-time web servers
HEIC: Lossy re-encode only (generational quality loss; cannot reconstruct original JPEG)

Fine Detail & Textures

JPEG XL: Retains sharp text, fine lines, subtle textures, and film grain
HEIC: Video-derived coding (HEVC / X.265). Tends to smooth out high frequences and smudge grain.

Software Encoding Speed

JPEG XL: Extremely fast; highly parallelized SIMD architecture
HEIC: Exceptionally slow and computationally expensive in software

Software Decoding Speed

JPEG XL: Fast, lightweight multi-threaded CPU decoding
HEIC: Heavy CPU overhead (without using hardware acceleration)

Progressive Rendering:

JPEG XL: True progressive decode; smart saliency algorithm to focus bandwidth on critical details first.
HEIC: None; full file must be received and decoded before display

Lossless Compression
JPEG XL: Dedicated modular mode; vastly outperforms PNG and WebP
HEIC: Ill-suited; you basically have to try to do lossless compression with an inherently-lossy format

Bit Depth & HDR

JPEG XL: Native HDR; up to 32-bit floating point per channel
HEIC: Typically capped at 10-bit or 12-bit integer

Max Dimensions & Scaling

JPEG XL: Up to 1B x 1B; efficient viewport/crop loading without full decoding
HEIC: v5.2, 4096x2160; v6.2: 8192x4320; a tiling hack allows up to 16384 x 16384

Channels & Color Spaces:

JPEG XL: Arbitrary color spaces (hyperspectral); unlimited extra channels (alpha, depth, thermal, masks, CMYK, etc)
HEIC: Rigid container; limited auxiliary channels and standard video color spaces

HEIC advantages:

Rollout / acceleration:

JPEG XL: no dedicated hardware acceleration (thankfully, it's not as important because it's so much more efficient). Software adoption still rolling out.
HEIC: Hardware silicon (ASICs). Default capture format on modern iOS/Android; native capture in Sony, Canon, and Nikon cameras

Video:

JPEG XL: Supports animations (GIF/APNG replacement), with some optimizations** (it's not just a series of stills), but lacks the full set of optimizations that a proper video codec has.
HEIC: Container naively pairs full HEVC video tracks and audio with video.

** - JPEG XL can store up to 4 reference frames in a buffer, with multiple blending modes from the references (add, replace, multiply, etc); has subframe bounding boxes ("dirty rectangles") for when only part of a frame changes; invisible frames; modular deltas (differences between frames); etc. However, it lacks motion vectors (e.g. detecting a feature drifting across a scene and simply having to encode "move it" (followed by any needed deltas). So it's great for "GIFs", but if you wanted to encode a full movie, it'd be significantly larger than e.g. HEVC.

I just looked at the paper and you are right. Thanks for looking.

The male and the female one are basically suggesting visually that the man is competent and the woman is not. I mean even only the cloth pattern of the skirt that "she" is wearing is atrocious. It is like they made the "woman" intentionally unattractive while they made the "man" average. ("She" is also blonde, of course, while "he" has dark hair.) I have no idea whether this was actually done intentionally, but if I were to design a study to find this type of bias, that is what I would use. I wonder whether they did something similar in the text interfaces.

I also noticed that their names are Johanna and Johan, i.e. a proper male name for "him" and a feminized male name for "her".

So, after all, my first take of "likely methodical error" was correct.

The study is flawed in another way as well because "everybody knows" that women earn less than men. (Usually that is "proven" with the unadjusted wage gap, which is completely bogus and essentially a lie.) Hence the study participants just conformed to what they thought were the facts of the matter. Makes the results completely meaningless.

And the other thing is that "everybody knows" there is a gender pay gap. This is only true for the unadjusted gap, but most people do not know what that means. Once you adjust for factors like experience and field-of-work the difference is below the margin of error, i.e. small and we do not even know whether it is real. We essentially have same wages for the same work and qualification and had it for a while. That some parts of some gender groups decide to not do the same work or invest less or more time into getting experience is a separate problem and not a problem of payment.

Hence the people in this "experiment" were just conforming to what they thought was expected of them. That makes the whole thing bogus. Obviously wages in the real world are determined by a different procedure and harping on about women and men not behaving exactly the same (as a group) is just one thing: counterproductive.

Well, what can an LLM do? It has no understanding or insight, so it cannot see whether something is a good solution or not. Obviously, a smart human would not even have looked for a "solution" like that, because it is obvious crap.

Are you serious? Because that is the dumbest shit I have read in a long, long time.

Since you do not even seem to know the very basics, here is a starting point: https://ancillary-proxy.atarimworker.io?url=https%3A%2F%2Fen.wikipedia.org%2Fwiki%2F...

No. Or rather not to any meaningful degree or depth. They can do very shallow reasoning only and after a few steps into deeper reasoning chains, only noise is left. Statistical AI is completely unsuitable for automated deduction. It is not even remotely comparable to what a smart human or even the very limited automatic theorem proving systems can do.

Yes. While this may be hard to understand for non-mathematicians, the data-theft angle could not be any more obvious. All LLM type AI can to is put puzzle pieces together. Somebody else needs to create those. And if they were unpublished, then they were stolen.

It does not actually make a useful contribution. It does damage. And the allegation of most of this stuff essentially being stolen are very likely true.

I am with Roger Penrose on this: Quantum theory is wrong. Saying it is "incomplete" is just a more polite form of saying it is wrong. Now, you do not abandon theories you know are wrong unless you have better ones. But you always remember they are wrong and you are careful not to make sweeping predictions based on them.

Yes. It takes a certain level of sophistication to overcome this effect. Some people have it. Many do not and for some of those it is completely out of reach.

You cannot force people to be fair or respectful or "good". You can only refuse to associate with those that are not.

There is also another aspect that makes this experiment bogus: Ordinary people do not decide about the pay of others in actual reality. It is a lot more complicated. This stinks of somebody trying very hard to find a specific problem.