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.
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.
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.
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
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.
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.
> 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".
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.
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.
-- 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.
> 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!