haldujai

61 ms·

691.
by haldujai 4y ago
Is it just me or does anyone else cringe when they read "is/is not all you need" in the title of an AI-related paper? Also, what is "SOTA" for review? There isn't exactly a benchmark to compare... In terms of c
692.
by haldujai 4y ago
What is your desired objective and what are you going to train on? There are plenty of publicly available model checkpoints so you don’t have to start from scratch with large GPU clusters. There’s no point in repeating the pretraining that
693.
by haldujai 4y ago
If you specifically mean a general LLM trained on a general language corpus with instruction finetuning this is correct. Fortunately very few real world use cases need to be this general. If you are training a LLM on a domain specific corpu
694.
by haldujai 4y ago
If you can’t fit the model on your resources you can leverage DeepSpeed’s ZeRO-offload which will let you train GPT2 on a single V100 (32gb). Alternatively, if you’re researching (with the caveat that you have to either publish, open source
695.
by haldujai 4y ago
Thank you for posting this. I’m not sure why but it seems like very few of the ChatGPT threads I’ve read across on Twitter/HN showing “very accurate and intelligent” results never experiment with ablation which is very important in pro
696.
by haldujai 4y ago
> If it were so automatic and repetitive, it would be easy to pass the information to the next medic and have normal length shifts. Ward call for residents generally works like this: I have an inpatient list of 15-20 patients I’m cover
697.
by haldujai 4y ago
Not sure what this has to do with resident/physician work hours. Nursing errors (i.e. administering the wrong dose) can certainly kill people. They also don’t / shouldn’t work 24 hours shifts (infrequently a nurse might work a dou
698.
by haldujai 4y ago
Doesn’t AWS offer this with SageMaker? Or huggingface auto train.
699.
by haldujai 4y ago
It’s essentially unheard of to have someone die because a resident made a mistake on call. On-call medicine is so rote as to not require much, if any, thinking. Ward medicine is far less intellectually challenging than teaching. Patients wh
700.
by haldujai 4y ago
This number includes level 2/step-down ICUs inclusive of regional/community hospital. This is not the bed count of units capable of having cardiac support or prolonged ventilation. I can’t readily find the OECD figure but if you l
701.
by haldujai 4y ago
It’s probably because we don’t have enough ICU (or step down beds) in Canadian hospitals than the fear of litigation in the US. Canada’s capacity is amongst the least in G20 nations. A lot of patients we manage on the ward or step downs (i.
702.
by haldujai 4y ago
As a physician, shift length is honestly a red herring. As much as I hated doing 24-28 hour shifts on inpatient services, continuity of care does matter and errors do occur in handover. You have to keep in mind that medicine between 12am an
703.
by haldujai 4y ago
I want to clarify two points given the language used in your response: 1. I used the possessive “my” in reference to nursing staff for simplicity in writing and clarity to the reader rather than to indicate ownership, we are on a team. This
704.
by haldujai 4y ago
I’m not sure where you’re getting this from. I / my nurses silence alarms at literally every hospital I’ve ever worked at (granted they’re temporary silences by design so you have to hit silence q1h/q30mins depending on the alarm)
705.
by haldujai 4y ago
A different angle that helped me understand your description better (graphic warning: scared children voices in the background audio) https://twitter.com/jasonwhitely/status/1591541681798668291?...
706.
by haldujai 4y ago
> The on chain activity makes this look like an individual who did not prepare extensively before doing this or doesn't have the skills to use automation/operational best practices This may be an understatement of their skill
707.
Matt Mochary CEO Coaching Curriculum(docs.google.com)
2 points by haldujai 4y ago|0 comments
708.
by haldujai 4y ago
“Just don't eat eggs and pray.” If you avoid smoking, have a healthy diet, get adequate exercise and consequently maintain a healthy weight you’ll do far more to prognosticate your MI risk than a calcium score.
709.
by haldujai 4y ago
Here are some resources to read. https://europepmc.org/article/nbk/nbk82767#_vaphysical_s5_ https://www.cfp.ca/content/cfp/63/11/824.full.pdf https://www.bmj.com&#x2
710.
by haldujai 4y ago
Suggesting we perform niche lab examinations annually on the entire population is not scientific research, that’s jumping straight into a screening program. Such an initiative would require more lab resources than currently exists in the US
711.
by haldujai 4y ago
An annual physical exam isn’t really powered to detect cancer in the modern world, it’s a relic from the past. If you develop cancer in your 30s (sadly too many people) a physical exam or blood test almost certainly wouldn’t have made the d
712.
by haldujai 4y ago
A billion dollars is not enough for what’s being proposed here.
713.
by haldujai 4y ago
It would be nice to have this information to see if any trends are discovered (all of this is theoretical, I question whether measuring excreted compounds rather than stores is is relevant but that’s a separate point) however it is unclear
714.
by haldujai 4y ago
It’s not. In radiology we are careful when we scan a fetus, especially with Doppler mode, because of the tissue heating effect.
715.
by haldujai 4y ago
Findings on imaging have value in specific clinical contexts. There are a lot of things you can see that will take you down a rabbit hole of further investigations and turn out to be nothing. We perform fetal ultrasonography which excludes
716.
by haldujai 4y ago
I’m not sure how safe that would be. I’m a radiologist and we practice with a principle called “as low as reasonably achievable” to minimize harm. While ultrasound is not ionizing radiation it does cause tissue heating. 24/7 ultrasound
717.
by haldujai 4y ago
Radiation oncologist*, this is far outside my (radiologist) scope!
718.
by haldujai 4y ago
This is incorrect. I’ve treated hypokalemia in thousands of patients. It takes days with oral routes. Also, a BP of 160 is not dangerously high. Particularly if you are young, healthy or known to the doctor as it is most likely white coat.
719.
by haldujai 4y ago
Unfortunately you are confusing “medical standard of care” with the “standards of medical care”. “The most common legal definition of standard of care is how similarly qualified practitioners would have managed the patient’s care under the
720.
by haldujai 4y ago
I can’t speak for the UK but 70+ percent of the US population lives in urban areas with over half living in the 48 largest regions. Therefore your conclusion that most individuals experience similar interactions with a “poor reputation sing

More ›