gwerbret

The article was written, from soup to nuts, by AI. I suspect the person who threw his byline on it barely glanced at the text.

Unfortunately, the equipment and reagent vendors in the very unregulated life sciences/biomedical research world constitute a racket. Serial buyouts over the last 10 - 15 years have led to a ridiculous degree of consolidation -- the sort that wouldn't fly in a regulated industry, or even one in which regulators are paying any attention at all -- so it's now dominated by two players: Merck (through its MilliporeSigma arm), and Thermo Fisher. The existence of this cartel means that they can essentially get away with murder, both from a fraud angle (which is exactly what the Western blot manipulation is) and by fixing prices to whatever degree they please.

Also unfortunately, biomedical scientists are not known for their tendency to collaborate to face a mutual enemy (the mild pushback against the Elsevier/Springer Nature publishing cartel has come less from scientists and more from university systems whose libraries have to foot most of the bills). From their perspective, it's "What am I gonna do? Raise my own antibodies? Start blowing my own glassware?" So they grimace and bear it.

For reference, here's how their workflow for research antibodies goes (and it's been like this for decades):

1. Produce an antibody the research world needs. Do no QA, that's expensive and unnecessary.

2. Claim with usually no evidence (and apparently by forging evidence) that the antibody works in certain applications.

3. Let researchers buy the antibody and do your QA for you. Even if the antibody doesn't work, only a tiny percentage of buyers will go to the effort of getting a refund.

4. Profit. Keep selling the antibody even when the rare scientist with time on their hands demonstrates beyond doubt that your antibody clearly doesn't work.

5. When sales start drying up because enough people are catching up to the scam, discontinue the antibody. Give no explanation.

6. Repeat from 1.

NMDA receptor encephalitis is usually associated with a particular ovarian tumor, so the first thing I did on seeing this article was to check if the author is male or female (he's male). It is the habit of certain cancers to present with bizarre symptoms (so-called paraneoplastic syndromes) including psychosis as in this case, and often it can be months before someone thinks to look for cancer. I'm glad the author's okay.

Can you elaborate? What's unique about the nations for which democracy isn't the best form of government? What's unique about their people?

Not the person you responded to, but functional democracies require either an already-established functioning government, which can be efficiently perpetuated and controlled by elected leaders (think the United Kingdom); or a non-functioning government which can be effectively reformed and then perpetuated by elected leaders (think Taiwan). In both cases, functional democracies also require an electorate who can be brought to work towards a nation's major development goals.

There are countries for which none of these criteria can be reasonably met in the context of a democracy. One reason is if the electoral processes are so corrupt that no one competent is actually elected or, if they are accidentally competent, are too busy working towards their own and their cronies' ends to be effective leaders. The second is if there is underlying social strife which prevents people from working collaboratively towards nation building.

India fails in both counts: the corruption at all levels of government is nothing short of legendary, and the country as a whole is comprised of very diverse peoples who, historically, have had little reason to work together. Many African countries, rather tragically, are in the same boat: during the colonial era, "countries" were almost randomly assembled out of groups of people who historically had almost nothing in common. When the colonial powers left, they typically left nothing behind -- no knowledgeable and experienced administrators, no established universal education, and little or no social infrastructure. The people were then left to reinvent government from scratch, and the "country" more often than not was actually five separate nations of people who hated each other.

In sum, democracy is sort of an advanced form of government which, when introduced, really does need a somewhat coherent nation to already exist (in more cases) for it to work well. An autocratic or authoritarian government is usually the on-ramp, so long as it's reasonably functional and stable for long enough. Wherever democracy has persistently failed to take off, it's invariably a place in which the underlying foundation didn't exist to begin with.

Well, mainly in the fact that Anna's has several orders of magnitude more books, and includes research publications and more, ah, contemporary materials to boot.

I love Project Gutenberg, don't get me wrong... but frankly, Anna's is better.

Methinks someone's experimenting with a botnet -- "can we bypass HN's checks and balances to get an entirely-irrelevant topic to the front page, and keep it there for x hours?"

Somewhat ironically, he'd spent the last years of his life working on prolonging life [1], and was selling a $25,000 "proactive healthcare service" consultation to anyone who could afford it [2].

1: The company's website, humanlongevity dot com, seems to have been compromised, and as "captcha" will try to have you install a Trojan. So here's the Wikipedia page instead: https://en.wikipedia.org/wiki/Human_Longevity

2: https://fortune.com/2017/02/21/craig-venter-human-longevity/

In case anyone wondered, the title is a play on the Isaac Asimov book "The End of Eternity": https://en.wikipedia.org/wiki/The_End_of_Eternity

You've had enough arguments with people in both this thread and the previous that I'm pretty sure you understand what the issue is with your use of the word "free".

What you are offering is NOT a free tool -- it is a demo, for a tool for which you are charging $12/month. No reasonable person would interpret a grand total of 3 exports as enough to justify calling this a "free" tool.

This is to say nothing of your violation of AGPL on the use of MuPDF, which has been pointed out here and elsewhere.

But of course, you're free to Show HN a paid product; just kindly don't insult our collective intelligences in the process.

As do most of the associated comments. I think we're surrounded by bots.

I really wish those offering speech-to-text models provided transcription benchmarks specific to particular fields of endeavor. I imagine performance would vary wildly when using jargon peculiar to software development, medicine, physics, and law, as compared to everyday speech. Considering that "enterprise" use is often specialized or sub-specialized, it seems like they're leaving money on Dragon's table by not catering to any of those needs.

The actual paper (open access): https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/202...

These are the important bits for the non medical folks

Also significantly: "vaccinated individuals consistently had a lower risk of death, regardless of the cause."

A bit tangential and pedantic, but:

At the heart of the problem is the tendency for AI language models to confabulate, which means they may confidently generate a false output that is stated as being factual.

"Confabulate" is precisely the correct term; I don't know how we ended up settling on "hallucinate".

I'm happy to see the move. Codeberg is probably a more stable/long-term solution than SourceHut as the founder is slightly unhinged

What's this about?

Some thoughts regarding Mozilla's leadership.

Certain aspects of human nature, as they apply to the corporate world, can be acknowledged and understood, even if they're not excuses when they lead to the downfall of a prominent organization. When you give someone a big title, a dump truck full of cash, and a mandate to innovate, human nature dictates that most people will internalize the idea that "because I was given all this, I must be competent", even if they very obviously are not. Typically the outcome is a "bold plan forward" which is notable for lacking any actual clear solution to the company's main problems. In one example I know of, the CEO decided to pivot from an unrelated field towards launching a cryptocurrency, and cooked up a cartoonishly-dangerous marketing scheme to support the idea. One person ended up dying as a result, and the company then purged every mention of crypto from its website. (And yes, the company collapsed soon afterwards.)

While it's easy to blame the CEO with their oversized salary, the blame for such disasters doesn't just lie with them. After all, arguably the most important roles of the board are to hire a good CEO, ensure the CEO is actually performing as they should, and fire them if they're not. When politics, cronyism, or again, simple incompetence, lead the board to also fail at its job, you end up with the long, slow decline into obscurity we've seen so often in the tech world.

But Mozilla had a good run.

Interesting work. Some thoughts:

First, your business model isn't really clear, as what you've described so far sounds more like a research project than a go-to-market premise. Computational pathology is a crowded market, and the main players all have two things in common: access to huge numbers of labeled whole-slide images, and workflows designed to handle such images. Without the former, your project sounds like a non-starter, and given the latter, the idea you've pitched doesn't seem like an advantage. Notably, some of the existing models even have open weights (e.g. Prov-GigaPath, CTransPath).

Second, you've talked about using this approach to make diagnoses, but it's not clear exactly how this would be pitched as a market solution. The range of possible diagnoses is almost unlimited, so a useful model would need training data for everything (not possible). My understanding is that foundation models solve this problem by focusing on one or a few diagnoses in a restricted scope, e.g. prostate cancer in prostate core biopsies. The other approach is to screen for normal in clearly-defined settings, e.g. Pap smears, so that anything that isn't "normal" is flagged for manual review. Either approach, as you can see, demands a very different training and market positioning strategy.

Finally, do you have pathologists advising you, and have you done any sort of market analysis? Unless you're already a pathologist (and probably even if you were), I suspect that having both would be of immense value in deciding a go-forward plan.

All the best!

Heh...this is a shady study if I ever saw one.

-- Exactly 400 study participants recruited.

-- Exactly 193 of 200 participants completing the study in each group (which, for a study administered in a community setting, is an essentially impossibly-high completion rate).

-- No author disclosures -- in fact, no information about the authors whatsoever, other than their names.

-- No information on exposures, lifestyles, or other factors which invariably influence infection rates.

-- Inappropriate statistical methods, which focus very heavily on p values.

-- Only 3 authors, which for a randomized controlled trial involving hundreds of people in different settings with regular follow-up, seems rather unlikely.

A little bit off topic, but even after years of active interest, I'm still amazed by the complexity of the human immune system.

Imagine this: we are all born with a functional immune system which is pre-programmed with knowledge of what bacteria, viruses, and many parasites look like, so it can immediately deal with these without prior exposure. This is the innate immune system, and in many organisms is the only immune system.

On top of that, a database is created which consists of fragments of all our bodies' molecules. This database is used to train the adaptive immune system. The thymus will then present these molecules to new white (T) cells, and screen out the ones that recognize these "self" molecules. This is the adaptive immune system.

Still on top of that, there's another tier, because maybe 0.1% of T cells escape the first-pass screening. You now have a series of checks and balances which screen for these escaped cells outside the thymus, and either reduce their functioning or eliminate them entirely. This is peripheral tolerance (what the Nobel prize in medicine was awarded for this year).

And when there's an actual infection, this system is able to spin up a few VMs, run a large bug-search model, and create a pool of tailor-made antibodies and T cells specific to the new bug, which in most cases are enough to deal with the infection.

So when all is said and done, and the system is trained and working as expected, you now have an immune platform which, on top of recognizing all its own molecules, can also recognize pathogens, including differentiating disease-causing ones from the benign ones; can also deal calmly with the enormous diversity of things we put in our mouths, noses, and other orifices; and in most cases doesn't actually go rogue.

But sometimes, it can be overcome by peanuts.

From one of the reviews:

You also pay 15 dollars after the first 60 minutes [for] another 15 min.

Really? 1-900-CALL-SANTA, only $1 a minute, must be under 18 and have stolen your parent's credit card, no refunds whatsoever? Merry Christmas to you, too!

Thank you. This is the sort of contextual overview that should really float to the top of discussions like this.

Q: how does Mozilla plan to diversify revenue away from Google's Search engine Deal in future? with Mozilla currently shutting down Pocket, Fakespot and other potential revenue sources?

Anthony Enzor-DeMeo, Firefox General Manager: Right now we make money through search and ads. Advertising helps sustain Mozilla’s mission by showing the industry that ads can work with privacy. Most people think of search as just Google; but we have 50+ search partnership deals around the world. We believe AI will present new revenue opportunities as well as some new tab experiences we are working on. We are exploring creative ways to drive different types of partnerships for revenue (widgets, feeds, etc).

So Google, ads, and throwing things at the wall and seeing what sticks.

This study's pretty wild -- but this approach has a major downside that they only mentioned in passing in the actual report in the New England Journal of Medicine (paywalled, unfortunately).

To gene-edit these cells, they had to use a lentivirus vector -- a (limited form of a) class of viruses that notably includes HIV. These viral vectors work by splicing themselves into random places in the host cell's DNA. Which is fine, except that there's a non-zero chance that in the process, the virus will initiate a cancer.

When you combine that with a cell deliberately engineered to hide from the immune system, you have the ticket to a very bad time.

What is the source that the healthcare costs are growing "exponentially" and are outstripping growth in GDP?

See here, in particular the first figure: https://www.cihi.ca/en/national-health-expenditure-trends-20...

And here (slightly dated, but still valid): https://www.fraserinstitute.org/studies/sustainability-of-he...

I live in Canada and have not seen any such messaging that you have said.

Here's a not-particularly-subtle example: https://www.cbc.ca/news/canada/new-brunswick/maid-medical-as...

There are lots of examples of MAID being pushed upon people that do have other options and made to feel like it's the only one.

I'm not really surprised. It looks like Canada's healthcare costs are growing exponentially, and are outstripping growth in GDP. These costs are mostly driven by hospitalizations. If a government can carefully promote the message that hospitalization means suffering, suffering is hard, a life with suffering is not worth living, and that relief is quick and easy, then a route is charted to a reduction in healthcare expense. It would certainly help if the large physician organizations are on board, and the nation's major broadcasters lean into euthanasia-friendly messaging.

There is a simple test the public can use for any scientific model: does it make accurate predictions, or not? You don't need to understand how a model works to test that.

It's quite obvious from your position on this matter that you're not a practicing scientist, so it's very unfortunate that your position is so assertive, as it's mostly wrong.

To understand the predictions, as it were, you do have to understand the experiments; if you don't, you have no way of knowing if the predictions actually match the outcomes. Most publications involve some form of hypothesis-prediction-experiment-result profile, and it is the training and expertise (and corroboration by other experiments, and time) that help determine which of those papers establish new science, and which ones go out with last week's trash. The findings in these areas are seldom accessible until the field is very advanced and/or in practical use, as with the example of GPS you gave elsewhere.

The biggest problem I see with "establishment" science today is that it doesn't work this way. There is no mechanism for having an independent record that the public can access of predictions vs. reality.

There is; it's called a textbook.

By the tenor of your response, I assume you understood what I meant in this very-non-medical of forums, which means you understood why the paper's authors themselves chose not to call their study a trial (even though they registered it as a "clinical trial", as is necessary for any clinical study in humans involving a treatment intervention). Which leaves me wondering about the purpose of your response.

Some important things not mentioned in this press release (not to detract from the idea of new treatment approaches of any sort):

- All patients had their tumors surgically removed before they were started on treatment. Thus the trial wasn't testing cure so much as delay of recurrence.

- These were very superficial tumors, meaning they were growing on the very surface of the inner bladder, just like skin tags. These aren't the ones that kill people. Patients with superficial bladder cancer who don't respond to BCG can be treated for quite a while just by having the tumors surgically removed whenever they recur (using a minimally-invasive procedure known as a transurethral resection of bladder tumors, TURBT).

- Fun with words: the press release called this a clinical trial, but it's not -- it has no controls, no real statistics, no randomization, none of the things that make up the usual standard in medicine. The authors of the paper call it a "study", which is basically a research experiment. They don't use the word "trial" at all in the paper.

Having said all that, I still look forward to seeing a proper trial.

Edit: wordsmithing.

Used Tesla car prices are now down 4.59% year-over-year

Their graph shows a drop from $33k to $28k since August '24, or about 18%. Did they leave a factor of 4 behind somewhere?