holub008

Has anyone found TPM/RPM limits on Nova? Either they aren't limited, or the quotas haven't been published yet: https://docs.aws.amazon.com/general/latest/gr/bedrock.html#l...

The IWPR link from above (with data from the US BLS) suggests that the 320K women in construction trades represent 4% of the total field, implying that there are about 8 million total construction trade jobs in the US. I'm not sure how many construction workers hold 4 year degrees, but it seems plausible this would explain a large portion of the gap.

Crazy, I wonder what AWS's margin is on these unauthorized requests? Questionable incentives.

Seconded, only for our international contractors.

Sorry about that! We had a high volume of applicants and needed to take the listing down for sanity. Please send along a resume to karl.holub at nested-knowledge.com

Nested Knowledge | Machine Learning Engineer | Remote | https://about.nested-knowledge.com/

Nested Knowledge is building the future of Systematic Literature Review (SLR) by consolidating every step of the review process (search, screening, extraction, synthesis) in an updatable, collaborative environment. NK primarily solves biomedical SLR, and our users come from a variety of backgrounds, including academia, insurers, med device / pharma, and regulatory. To date, 2 million studies have been reviewed in our software.

We want you to build models taking our Robot Screener and Smart Tagging products to the next level and design automations in new areas, like critical appraisal. Our team is small, and you would be our first AI specialist, setting standards and frameworks in place. We value independence, creativity, and a focus on end-user value and experience over technology.

https://wellfound.com/jobs/2866132-machine-learning-engineer

From a different direction, it was taught in one section of my university's data structures course. We were guinea pigs (https://forum.dlang.org/post/km96ho$2grm$1@digitalmars.com), but I recall the experience being fun and students largely enjoying D. Transitioning to C++/Java in later coursework and professionally was a breeze.

Good point, but headshots are essentially required in arenas other than job hunting. For a startup, it would be unprofessional to pitch investors without headshots of the founding team in the deck. Even in hiring, I'm willing to bet most hiring managers look candidates up on LinkedIn where they will see a headshot.

Does this dilute the headshot currency in the professional marketplace? If anyone can generate a flawless, idealized photo of themself, surely it reduces the signal headshots once provided: that you have the means to pay a photographer & are willing to invest the time / give a damn about appearing professional.

It might also damage the sense of identity they provide, if the generated images wander too far from reality of the source images.

Right, MEDLINE's "tagging" system is MeSH, which is a large controlled vocabulary. MEDLINE does contain bibliographic data + journal keywords, but its real value add is MeSH, which is used for search, related publication identification, etc. in PubMed.

https://en.wikipedia.org/wiki/Medical_Subject_Headings

Content tagging in online systems has existed since at least the 1970s, with the earliest example I can think of being MEDLINE https://en.wikipedia.org/wiki/MEDLINE

Interestingly, the original piece is a later Beethoven Sonata, Op. 31 No. 3. The model has its styles down! https://youtu.be/P-Q5aBAw-T4?t=78

How else would you like them to state that without being overly verbose?

"Our results suggest that some sexually transmitted parasites, such as T. gondii, may be correlated with appearance and behavior of the human host."

I appreciate your viewpoint. I would counter it by saying that there are two sources of uncertainty here: choice of model & sampling variance. It's my opinion that in scientific writing, one should be precise with which source of uncertainty they are guarding. If I'm allowed to group these together, why can't I make a similar statement of causation of any old spurious correlation - when obviously my model is bad?

Considering this example again - isn't it arbitrary that the authors get to choose which hypothesis (among many, like attractive people being predisposed to own cats) they get to claim "may" be demonstrated?

Similar line of discussion: https://statmodeling.stat.columbia.edu/2015/04/04/thinking-p...

I was directly quoting the article and am not a domain expert. It is interesting to see some commenters here state that it isn't an STD. It appears this is an open question.

https://academic.oup.com/aje/article/190/3/386/5905708 https://pubmed.ncbi.nlm.nih.gov/24986706/

I wasn't implying infection was occurring during sex, but rather by being in the environment (many people stay around partners for more than sex).

Perhaps it's because I don't frequent the literature, but I interpret that as "T Gondii. may [provided by our level of statistical certainty in MANCOVA] produce changes". Otherwise, what's the point of performing a statistical analysis? Moreover what's the value of any assertion if it can be guarded by an unbounded-uncertainty keyword "may".

How about correlation between attractiveness and owning cats? Attractiveness and having a larger number of partners, which increases odds of exposure to cats / T. Gondii? These hypotheses seem plausible enough that I believe the authors are wrong in claiming causality.

Seriously, has the bar sunk this low? I was certain the authors would express caution in interpreting causality, but alas: "Our results suggest that some sexually transmitted parasites, such as T. gondii, may produce changes in the appearance and behavior of the human host".

Close to nothing of what makes science actually work is published as text on the web

Unless there's some nuance I missed, I immensely disagree with this statement.

I'm currently in the biomedical literature review space, and I appreciate the detailed insights. I wonder if the author considered that literature review is used in a wide variety of domains outside pharma/drug discovery (where I perceived their efforts were focused). Regulatory monitoring/reporting, hospital guideline generation, etc.

This is a billion dollar industry, and I couldn't agree more that it's technologically underdeveloped. I do not agree that AI-based extraction is the solution, at least in the near-term. The formal methodologies used by reviewers/meta-analysts: search strategy generation, lit search, screening, extraction, critical appraisal, synthesis/statistical analysis, are IMO more nuanced than an AI can capture. They require human input or review. My business is betting on this premise :)

I was surprised by North Dakota as well. I believe that bright spot corresponds to oil fields.

https://en.wikipedia.org/wiki/Bakken_Formation

Thanks for digging into this- what a cool adventure!

In your research, were you able to find GPS coordinates of the purported island? I'd be interested in taking a trip out to search myself, but doing a grid search of the whole island sounds challenging based on your report.

Upon a little more research, in an NPR interview [1], Dickey does not claim to have exhaustively searched the island / proved non-existence:

DICKEY: We hiked in as far as we could to try to find the center of the island, was unable to find Moose Flats or Moose Boulder. Who knows; it could have been there, but highly unlikely.

[1] https://www.npr.org/2020/03/13/815546895/the-lonely-non-isla...

Agreed on both of your conclusions. Looking at the island's topography [1], I'd be hard-pressed to believe there's a pond anywhere on that island, excepting maybe the northeast corner. But, their story doesn't add up. The timeline of the article states they paddled to Ryan Island and then it immediately jumps to them getting lost on the paddle back to camp. The whole story is about verifying the existence of this "island", so why is this crux of their journey not mentioned?!

I frequently canoe and camp around northern Minnesota and Lake Superior. I do not understand how they could have possibly traveled a total of 18 miles from Malone Bay to Ryan Island, as the article states. At most, it would be a 1 mile portage and a 5 mile paddle. Getting lost and adding 12 miles seems very unlikely, as you mention, due to all the navigable landmarks in Siskiwit Lake. And that's ignoring GPS.

My guess is that they are not strong paddlers and/or navigators, got to the island already tired, saw the amount of bushwacking that would be required to explore the island, and bailed.

For the record, I think this is really cool. Something about the story, as presented, doesn't sit right though.

[1] https://www.openstreetmap.org/#map=17/48.00956/-88.76993&lay...

There's a bunch of comments in this thread from people who did not read even the abstract of the study.

Those 3 data points were used in a graphic, which is not relevant to the actual experiment performed. Their experiment focused on cohorts of "parents who were (just) entitled to the new paternity leave" and "(just) ineligible parents". This design provides quasi-randomization so that any prior "downward trend" is not relevant, and clearly there are more than "3 data points" at play.