the irritating thing about LLM generated papers like these is that they're wrong but are generated using LLMs that are capable enough to bury the absurd claim pretty deep in there.

Analyze it using an LLM. Claude was pretty ruthless about this one.

Yeah, for me Claude identified the phrase "this holds with probability 1 over random weight matrices since the null space has dimension"

Treating trained weights as random for the purpose of a proof is immediately discrediting for a paper to me.

"This holds for almost all matrices" is actually something you'd want to know if we're talking about probabilities, no?

sure but it seems spiritually wrong to use an LLM to debug a slop paper. Who knows, maybe claude generated it in the first place?

In fairness, real-world researchers are expert at selective emphasis too.

this paper looks AI generated to me... I mean, there's no experiments to go along with it.

I mean I get your argument but it feels like one should adjust for wage growth instead. One labor unit of value converts to a shittier backpack.

The other side of that coin is that someone whose units of labor demand less value can still get into the market.

see terry pratchett's boots theory of economic fairness. They'll get into the market with something that costs more long term...

Exactly, and many who could have saved up for the original quality product are also lured into getting the crappier one that doesn't last as long - often without being aware of that deal.

for making research grade devices you barely need a cleanroom

You just need unitarity.

Unitarity means that information (about quantum states) is not lost, despite it appearing otherwise after a measurement. The Many-Worlds interpretation seems to be the simplest way to explain where this information has gone.

There are NO alternatives. There's nothing else that stays liquid at 4 K and absolutely nothing else comes close.

> There are NO alternatives.

We use a lot in our MR scanners.

The tech is changing and magnets are using far far less.

Super-conduction at higher temperatures has made progress too.

So while you are right that nothing else stays liquid at those temps, we won’t be needing nearly as much helium in radiology in the next few years.

The new generation use something like 700ml of helium, where the standard was hundreds of litres. https://magneticsmag.com/siemens-healthineers-gets-fda-clear...

> So while you are right that nothing else stays liquid at those temps, we won’t be needing nearly as much helium in radiology in the next few years.

How many loans for MRI machines that require helium haven't been paid back yet?

I know of a few.

They use a lot when installed, but rarely need top-ups.

They are shipped full of helium and chilled, but aren’t ramped up (so aren’t superconducting magnets until after commissioning).

The article itself spells out several alternatives to buying continuous amounts of Helium: high temperature semiconductors and zero boil-off systems that don't require a continual supply.

All these "we're going to run out" stories pretend that engineering cannot adapt to changing cost structures, which is just total nonsense.

Sure, there is nothing that can be directly substituted for how we use Helium today, but clearly we're using Helium inefficiently today and the answer is that once markets force us to change, we will find more efficient ways.

The article also points out several cases where this isn't possible

2d materials are so awful to work with but keep yielding these stunningly beautiful results so physicists must persist.

Pessimism of the intellect, optimism of the will

The power of quantum computing is constructing the solution to a problem out of an interference pattern. Classical probabilities don’t interfere, but quantum probabilities do. Loosely, quantum probabilities can be constructed to cancel, since their amplitudes can be negative.

Shor’s algorithm works on the quantum Fourier transform. The quantum Fourier transform works because you can pick a frequency out of a signal using a “test wave.” The test wave can select out the amplitude of interest because the information of the test wave constructively interferes, whereas every other frequency cancels. This is the interference effect that can only happen with complex/negative probability amplitudes.

The communication here is clear as mud. WHICH quantum systems? D-Wave? We know D-Wave is a joke!

The communication is in a superstate that has yet to collapse.