It's a modeling of language, it's not structurally anything like an LLM.

It is structurally exactly the same thing as an LLM...

It’s literally a trigram (character) language model. Check any NLP book from before 2015 or so.

LLMs have more stuff bolted onto them (embeddings, RLHF) but the autoregressive core is a direct descendent of that sort of language model.

So, LM vs LLM? :)

Isn't this pretty much why language models were invented?

Pasting something directly into the chat interface seems weird, but if you could somehow just see where P(token | context) falls off a cliff, that's a pretty good hint that your writing has problem.

Yeah, but for this use case you don't need Claude. You probably want a tuned lightweight small model that can run locally.

Even Haiku is massive overkill for this use case.

I've done it to check transcriptions of handwritten 19th century Quebec French. Is there a dedicated spell checker for that?

Ha! I’d never thought about it like that but…yeah.

I suspect another big part of it is that marketing and sales are relatively easy to measure and to scale.

You can hire one, two, or three new salespeople and expect that revenue will change more or less proportionately. Fixing (or ignoring) a handful issues doesn’t scale so smoothly—-there are jumps where the product suddenly seems much better/worse.

And I’m sure those companies also have “backlogs" due to limited labor/labor costs. There are always shelves to face, vehicles with deferred maintenance, and so on.

Obviously, there are limits: I’m not sure what my local grocery store or bus line would do with 100 new workers, but I have no doubt they could put a few people to work right away.

Well, but they aren't doing so right now, because labour has costs.

If your existing labour suddenly gets more productive, there's still a trade-off between cost and benefit to be made. And it would be a great coincidence if the optimal trade-off were exactly at the same headcount as the old trade-off under the old productivity figures.

It's more likely that the optimal trade-off for the business under the new conditions is at some other headcount.

I wouldn't say "called into question", as if the whole idea is bunk.

MRI is, in general, a lot harder than people often imagine. It uses complicated physics to measure convoluted physiological changes to indirectly measure brain activity, which is obviously stupifying involved--and then relate that to other, often complicated factors like behavior, lifestyle or disease state.

I think it's reasonably well-known that the BOLD response is complex and doesn't directly reflect "average" spiking activity. Some studies find that it's sensitive to the amount of synchrony (=more neurons firing together in time) rather than the rate. The paper you mention shows another dissociation: neurons can get more fuel by extracting oxygen more efficiently OR have having more overall oxygen to extract at the same rate. Thus, it's not noise, but it is complicated.

I'd be surprised if it were possible to directly measure muscle activity with millimetre wave radar. It looks like they're detecting motion, which is related to motor activity, of course.

EEG "spellers" c. 2000 required a cooperative participant who's actively engaged in a behavioural task: you attend to the letter/word you want to send and thereby produce a different response when it (vs the other letters) flash.

Implanted electrodes can do a lot better but it's still not something that will let someone "slurp" your thoughts out of you -- it'd be like subvocalizing them.

I do believe that signal deciphering part is somewhat solved. All the way up to real-time text stream. You can even go and run it yourself if you want to (e.g. https://github.com/CNN-for-EEG-classification/CNN-EEG). What I do struggle to find a proof of is signal gathering part. What do you think of this patent https://patents.google.com/patent/US3951134A/en ?

In case you were wondering - exactly this is what I've been listening to for 2.5 years every day 24/7 https://pastebin.com/raw/4AaidkHk it's AI generated slop in Latvian language, ran through voice synthesizer, filtered and attenuated down to hide imperfections. I do think authors have some sort of feedback mechanism whatever that may be. If not straight up "slurping thoughts", then perhaps overall emotional state changes, perhaps just heart rate - don't really have a way of telling. One thing I'm sure about is that it's external. I'm well familiar with sound, signal processing, programming. I'm also familiar with audio hallucinations - what I'm forced to experience is not that. Despite that such suffering is taking quite a toll - I don't believe it's schizophrenia (that's a whole new rabbit hole). Mainly because said disease is supposed to develop much earlier, because I've been living non-intoxicated and because whole voices thing started suddenly as per light switch (end of 2003 right when Russian Internal Revenue Service got hacked allegedly by Ukrainian hackers).

meant 2023

The man who mysteriously disappeared whilst swimming has a Naval Communications station and a pool complex named after him?! Amazing!

As an Aussie, it’s always amused me that we managed to name a public swimming pool after a Prime Minister who (at least as the official story goes) died whilst swimming.

The automated ones don't care, but it absolutely matters for the informal credit assignment process that actually runs academia.

I really wish we had a better way to "name" papers. Big clinical trials often have an acronym (often hilariously forced: "CXCessoR4"). That takes the emphasis off (one) lead author but it's implausibly hard to make up one for every research paper.

What "informal credit assignment"? It's automated and it runs entirely on quantitative data.

the one where i think of a particular piece of work, and i know who did it, then tell a student "oh, see if $author's group published anything else about this."

i'm not using software for this if this is off the top of my head, and it's the sort of thing that, at scale, hurts the forgotten author and their students

There’s a cute study demonstrating this effect by comparing career success in economics and psychology.

The author lists for economics papers are traditionally alphabetized, so more of your output will be known by your name if it occurs early in the alphabet. Abbie Ableson gets lots of mentions as "Ableson et al." while Zhang Zhu will almost always be relegated to the "et al". If name recognition matters, you’d expect successful academic economists to be clustered at the beginning of the alphabet—-and this appears to be true.

In most psychology journals, the author list is instead ordered by contribution/senority, and this effect disappears. https://www.aeaweb.org/articles?id=10.1257/08953300677652608...

Right, academics who deligate their entire intellectual life to GPT will be unaffected.

Right, and everyone else unaware of this made up "informal credit assignment process".

I don’t know that everyone would label it like that, but it’s inarguably true that success in academia comes from your reputation/name recognition.

Metrics are often attempts to formalize this but they’re not how most people actually make decisions: nobody is inviting seminar speakers or choosing collaborators because they have a high h-index. If anything, it goes the other way: name recognition gets you invited to speak or collaborate, which makes more people aware of your work, which boosts metrics.

That is false. The first thing everyone (at least everyone in CS---IDK about other fields) looks at are h-indexes, impact factors, number of papers per year, university rankings, and similar metrics. Researchers are most definitely selecting collaborators with a high h-index.

So we're talking about this woman's contribution. And you're talking about how the system is depriving her of recognition.

Do you see the inherent tension in what you're claiming vs the lived experience of everyone in this post (including you!)?

Cmon…We’re saying that a certain style of reference gives her less credit than might be due. Not none at all.

One paper doesn’t make a career (she wrote many dozens), it’s not always cited weirdly, and even if it is, some people may remember the coauthors (as they should).

But since you mention lived experience, I’ll add that I’ve certainly been asked if I’m "even aware" of results from co-authored papers where my name was listed second—-and I don’t think this is very uncommon experience.

You rediscovered normal saline!

The plain water hurts because it's causing cells to swell or even burst as water rushes into them to equalize the osmotic pressure. Adding a little bit of salt to that helps remove that pressure because the environment inside and outside of the cells are both equally salt.

Well, not so much rediscovered it as reasoned that that might be the cause of the pain and then confirmed it. The shocking part is that no doctor advises to use this after a tonsillectomy (or other kind of mouth or throat procedure).

My dad (also a doctor) called it a typical case of "the nurse's wisdom": the kind of quality of life interventions that typiqally get discovered by nurses and passed down orally, but never make it to official medical journals.

I don't think carprofen has been sold for human use in ages! Still available as a veterinary medicine though.

I wasn't sure how old the story or where the OP is. But yes it was pulled from human usage in most countries. I don't fully understand why, although I suspect it is primarily because ibuprofen is effective enough