In many people a momentary drop to the 80s or even below is not an emergency or anything close to it. Not saying that it is good.
Someone that is awake, sitting up, and struggling to breathe should be considered an emergency regardless of oxygen levels (and in this situation 80% would be very concerning).
EDIT: your comment is otherwise entirely correct, particularly at sea level.
One a friend and I hooked ourselves up to continuous pulse oximetry and had a contest to get the lowest recorded oxygen level. We tried everything we could think of, from just holding our breath to end-expiratory breath holding to hyperventilating to clear O2 (I used to do some recreational free-diving) beforehand to exercising (jumping jacks)...
Neither of us could get it below 98%, and this was at a mile of elevation (UNMH in Albuquerque).
It is pretty easy if you use the Wim Hof method. Breathe deep and fast for a few minutes, to the point where you get dizzy or feel weird sensations. Then exhale and stop breathing while being fully relaxed. I've done this while hooked to a pulse oximeter and it takes quite a while before O2 actually starts dropping (especially because the effect can be delayed in your limbs), but once it starts you'll pretty quickly run into the regime where a normal oxi will start an alarm because O2 is too low. You can even go below 85% without losing consciousness, because your limbs will desaturate faster than your brain. It's also not uncomfortable, because rising CO2 is what causes breathing reflex, but you dropped its levels far below the threshold by hyperventilating first.
Interesting. As I noted above (though typo O2 -> CO2), I used the same technique you describe, which I learned in free diving, and was not able to get below 98%, at altitude.
I'm quite sure no freediving instructor would ever teach you this particular method, because it is a surefire way to die underwater on your first attempt. Free diving breathing techniques usually revolve around lowering your heart rate, not lowering CO2. Wim Hof trainers will also tell you to never to use this method when near water.
As for the specifics that may have prevented you from doing what you wanted: If you breathe too shallow or too slow, you won't clear enough CO2. In freediving this is normal (even wanted), but for Wim Hof practice it means you didn't do it right. You really have to breathe so deep and fast that you enter an uncomfortable zone. It's not unlike physical exercise, except it's mostly mental.
> I'm quite sure no freediving instructor would ever teach you this particular method, because it is a surefire way to die underwater on your first attempt
Definitely no instructor involved, just a dumb 20 year-old living in Puerto Rico. It admittedly was dangerous, but I am living evidence that it was far from a "surefire" way to die. It was one of a hundred ways in which I put my life at risk during my 20s. ¯\_(ツ)_/¯
> As for the specifics that may have prevented you from doing what you wanted: If you breathe too shallow or too slow, you won't clear enough CO2.
I'm confident I was doing it sufficiently well to accomplish a longer period of breath holding than I otherwise would have been able to sustain, as evidenced by having done so (in addition to the usual symptoms of lightheadedness, confusion, loss of vision, near-syncope -- yes I agree quite uncomfortable). I know people on HN love to idolize Wim Hof, but in this context minute ventilation is not that difficult of a concept; I'm usually able to estimate the response in a paralyzed patient's PCO2 fairly well when making changes to their tidal volume and rate.
I didn't search for too long, but here's at least one relevant document, in which otherwise untrained subjects were able to achieve a substantial reduction in CO2 (17.4 vs 29.0) with a mere 15 seconds of hyperventilation, leading to an extra 23 seconds of breath holding prior to involuntary breathing moments. The peripheral O2 sat nadir in the hyperventilation group appears to have been identical to the non-hyperventilation group after the first trial (Fig 8b, looks like ~94%) and was only statistically significantly lower on trials 2-5: <https://pmc.ncbi.nlm.nih.gov/articles/PMC10363065/>
90 seconds of breath hold is still way too little to see oxygen drops in a finger pulse oxi. Also explains why you didn't die free-diving. Even with no training you can go much longer than that before risking a blackout (at least under normal circumstances). In fact training free-diving is all about CO2 tolerance and relaxation, so you won't even be able to store more O2. When I was reaching the 80s, I was doing 3+ minute breath holds. The first ~2 minutes my oxi stayed at full O2 sat with basically no change. In principle you should be able to induce a blackout yourself using this method without feeling an urgent need to breathe - if you do the prep stage hard enough. Just make sure you only do this when lying down in a safe position.
Can’t claim to know for sure, but I’d assume some kind of measurement limitations: the resolution and upper/lower limits of pulse ox are probably calibrated to some medical need, not to detect changes beyond what’s medically necessary
This was using equipment in the emergency department of our state's only L1 trauma center and comprehensive stroke center; I presume it was decent as far as medical monitoring equipment goes.
> Even now, I'll take a surgeon that's a complete jerk over a nice surgeon any day, because if they've got that job even as a jerk they've got to be good at their jobs.
This seems like an incredibly poor line of reasoning.
Hospitals are often desperate for surgeons. The poorly mannered ones are often deeply unsatisfied, angry at the grueling lives they've opted into, and the hospitals can't replace them. The market is not exactly at work here.
So why not order your own labs? I'm sure you can think of ways to get your own medications if you are sufficiently convinced that this is the best course of action for your health.
Arterial blood gas. Calcium score (may not count as a lab). Skin biopsy for cancer (does it count as a lab?). I'm unaware how to order my own troponin if I think I've had a heart attack (not that that's one I should DIY diagnosis). Prostate specific antigen.
Several of those are more procedures than labs. Of course you can't get someone to do a procedure on you for free. Arterial sticks and biopsies may have nontrivial risks (and commensurate liability risk for the performing provider).
PSA and troponin seem trivial to get. Did you look?
Unfortunately the training data is absolute garbage.
Diagnostic standards in (at least emergency, but I think other specialties) medicine are largely a joke -- ultimately it's often either autopsy or "expert consensus."
We get to bill more for more serious diagnoses. The amount of patients I see with a "stroke" or "heart attack" diagnosis that clearly had no such thing is truly wild.
We can be sued for tens of millions of dollars for missing a serious diagnosis, even if we know an alternative explanation is more likely.
If AI is able to beat an average doctor, it will be due to alleviating perverse incentives. But I can't imagine where we could get training data that would let it be any less of a fountain of garbage than many doctors.
Without a large amount of good training data, how could AI possibly be good at doctoring IRL?
You just get 1M doctors to wear body cams for a year. Now you have a model that has thousands of times your experience with patients, encyclopedic knowledge of every ailment including ones that never present in your geography, read all the latest papers, etc..
I don't understand how you think this doesn't win vs a human doctor.
This wouldn't solve the problem of diagnostic standards. Let's say you are a pediatrician and want to predict which kids with bronchiolitis will develop respiratory failure and need the ICU versus the ones who can go home. How do you determine from the body cams which kids had bronchiolitis in the first place? Bronchiolitis is a clinical diagnosis with symptoms that overlap with other respiratory illnesses such as asthma, bacterial pneumonia, croup, foreign body ingestion, etc.
you would have footage of the doctors diagnosing them. I don't understand what you're asking. The body cams have microphones too in case that wasn't clear.
In healthcare, HIPAA/GDPR equivalent would block this. Let's be realistic in our discussion; this is not the same as google buying up a library worth of books, scanning and destroying them
Other countries actually don't necessarily have a similar mix of ailments, median patient appearance and style of communication or even recommended course of action and most of the ones with more sophisticated medical care also have strict medical privacy laws. If you're genuinely unaware of this, I'm not sure you're in a position to be making "one year with a camera, how hard can it be" arguments...
(Where AI is likely to actually excel in medicine is parsing datasets that are much easier to do context free number crunching on than ER rooms, some of which physicians don't even have access to ...)
I think you're being silly if you think the amount of money at stake here, not the mention the health of billions of people is going to be stymied by privacy laws.
We have wildly heterogeneous data just within the US!
And again, how exactly is this interface going to work? How does the AI determine how hard to press on an abdomen, and where, and how does it press there once it has that information?
I'm still fairly new to local LLMs, spent some time setting up and testing a few Qwen3.6-35B-A3B models yesterday (mlx 4b and 8b, gguf Q4_K_M and Q4_K_XL I think).
Was impressed at how they ran on my 64G M4.
It looks like this new model is slightly "smarter" (based on the tables in TFA) but requires more VRAM. Is that it? The "dense" part being the big deal?
As 27B < 35B, should we expect some quantized models soon that will bring the VRAM requirement down?
that's not it. 35B-A3B is a "Mixture of Experts" model. Roughly, only ~3B parameters are active at a time. So, the actual computational requirements scale with this ~3B, rather than with the 35B (though you need high-bandwidth access to the full 35B layers though).
This model is a "dense" model. It will be much slower on macs. Concretely, on a M4 Pro, at Q6 gguf, it was ~9tok/s for me. 35-A3B (at Q4, with mlx, so not a fair comparison) was ~70 tok/s by comparison.
In general dedicated GPUs tend to do better with these kinds of "dense" models, though this becomes harder to judge when the GPU does not have enough VRAM to keep the model fully resident. For this model, I would expect if you have >=24GB VRAM you'd be fine, e.g. an NVIDIA {3,4,5}090-type thing.
I strongly disagree. If it's doing well enough for the owner then it's doing well enough. I don't understand how one can tell someone else that their computer is unacceptably slow for that other individual's personal use.
This is a really unfortunate move by Amazon. My next e-reader will be one that I own (instead of just rent).
Glad that I took the time to jailbreak and pause updates on my 2017 kindle paperwhite while I could.
I'd suggest cheap Android-based Chinese e-Ink e-readers if you want flexibility. My current one is a Bigme B6, which was for sale in my country a few months ago.
Their main advantage is providing access to all e-reading apps available on the Google Play Store, including Amazon's own Kindle app, as well as sideloaded ones such as KOReader.
On the downside, the battery life on those isn't as good as that of dedicated Kindles, Kobos, or other lightweight e-readers, but they still hold a charge for four or five days if one turns off their antennas, which is plenty of time to recharge them.
As for the ebooks themselves, I switched to purchasing from Kobo and other ebook stores. Some sell DRM-less ePubs, which is nice, while those that come with DRM can be easily liberated. And for the occasional Kindle-exclusive that is struck with (temporarily) unbreakable DRM, the Kindle app, although annoying, works well enough.
I'm pleased with OBOOK5. It runs Obook OS which is a Linux OS. Never nagged me to connect to WiFi or anything, I simply plugged a cable to transfer my local stuff.
Same here, I quite enjoy it. Plus there is open source software available, such as crosspoint. It’s easy to flash and an opus call away to change the behavior if you want something to work differently.
Yeah I flashed Crosspoint on it as soon as I got it.. Seeing the improvements it provided was partly what convinced me to buy the device.
I really appreciate that the company that makes the device has embraced the community firmware scene and even links directly to them from their website as a semi-blessed alternative to their official one.
EDIT: your comment is otherwise entirely correct, particularly at sea level.
reply