Steph Eaneff retweetledi
Today I learned that there is nothing - absolutely nothing on earth - that can prepare you for the gripping existential fear of seeing one of your git commit messages unexpectedly show up on the news
English
Steph Eaneff
112 posts
Steph Eaneff
@stepheaneff
Data scientist. Coffee lover. Policy enthusiast.
Katılım Nisan 2014
562 Takip Edilen72 Takipçiler
Steph Eaneff retweetledi
Steph Eaneff retweetledi
Data contributions are currently undervalued in AI.
We need more conferences and journals (like NeurIPS datasets and benchmarks & Nature Scientific Data) that give dataset curation the appreciation it deserves!
English
Steph Eaneff retweetledi
Check out @RebeccaKatz5 in Nature today talking about our work documenting the global history of outbreaks! nature.com/articles/d4158…
English
Steph Eaneff retweetledi
How to build a Global Burden of Climate Change Study:
📝 Equitable data sharing mechanisms to build transnational epi line datasets
📏 Detection and attribution methods from climate, applied to health outcomes
🌍 Scientific leadership from the frontlines of climate injustice
English
Steph Eaneff retweetledi
It sounds inconceivable, but it's true: scientists haven't updated their estimates of the total worldwide deaths caused by climate change for 20 years.
It's time to change that. That's why we're calling for a Global Burden of Climate Change Study.
journals.plos.org/climate/articl…
English
Steph Eaneff retweetledi
I finally made a website for my stats/DS/R artwork (thanks for dealing with my unruly ReadMe for so long, everyone😅).
I hope the site makes it easier to explore, find & use the artwork in your materials. This is where I'll add new stuff from now on.
allisonhorst.com
English
Steph Eaneff retweetledi
Mariame Kaba: “Hope is a discipline.”
Paul Farmer: “You fight the long defeat.”
James Stockdale: Combine “the need for absolute, unwavering faith that you can prevail” with “the discipline to begin by confronting the brutal facts.”
These are my touchstones. 8/
English
Steph Eaneff retweetledi
What % of #EHR text is directly duplicated from a prior note?
6 yrs
100+ mil notes
33 B words
192 B characters
Duplicated text ⬆️ from 33% in 2015 to 54% in 2020
Blocking copy forward is not the answer. We must reimagine #documentation
jamanetwork.com/journals/jaman…
English
Steph Eaneff retweetledi
Do you collect empirical data on wildlife disease prevalence out in nature? Are you interested in being part of @viralemergence's next big experiment in open data sharing? Drop the host-pathogen system you study (and where!) in the replies, and we might reach out 👀 Please RT!
English
Steph Eaneff retweetledi
court transcripts of lawyers questioning experts about statistical concepts is a highly underrated genre
English
Steph Eaneff retweetledi
Two open staff positions with our new viral emergence institute, based here with my group in Washington, D.C. at Georgetown University. Whether it's the science-of-team-science approaches or the SQL databases that get you up in the morning, consider applying:
The Verena Institute@viralemergence
PLEASE RT: We're hiring! Two full-time positions: • a DATA SCIENTIST expanding & curating our open data, developing software, and supporting research • a PROGRAM MANAGER for the Institute's entire activities, especially our education & training core Details below 👇
English
Steph Eaneff retweetledi
Pay attention: Big data loses to good data. “A survey of 250,000 respondents can produce an estimate of the population mean that is no more accurate than an estimate from a simple random sample of size 10… data quality matters more than data quantity” nature.com/articles/s4158…
English
Steph Eaneff retweetledi
For all other tools, we understand them as things that humans create, refine, adopt to perform some function. They extend our abilities. In claiming "superhuman" for AI, we are claiming that it does everything humans do and then some, and therein lies all kinds of problems.
English
Steph Eaneff retweetledi
A disproportionate number of discussions in clinical AI ethics involves technologies and use cases that don’t currently exist, at the expense of considering the ethics of technologies that are already widespread and affecting care for many people today.
Giada Pistilli@GiadaPistilli
Not only does this way of thinking hurt the philosophical field, but focusing on these sci-fi issues only perpetuates the collective panic that exists around these technologies while neglecting their actual risks. (8/11)
English
Steph Eaneff retweetledi
We don't have the luxury of not getting this right. Global access to adequate healthcare needs to be a part of these conversations—and that requires real funding and a hard look at the biases that have shaped these systems.
English
Steph Eaneff retweetledi
Building good quality datasets is orders of magnitude harder than building fancy or even state-of-the-art machine learning systems. Building #HornMT was a humbling experience. Thanks to our team of translators, @GebrekirstosG and @abel_aregawi.
12/n
English
Steph Eaneff retweetledi
New preprint! We built a database of the @WHO Disease Outbreak news, including all 2,789 reports from the first 25 years. It's both a global outbreak record and a window into WHO's operations. (We even indexed some data from 2019 that's currently lost!)
medrxiv.org/content/10.110…
English
Steph Eaneff retweetledi
Task design is such a crucial part of evaluation.
I genuinely think a lot of ML failures are just the result of ML methods being thrown at tasks that are completely inappropriate for ML to solve (ie. data with few examples, sparse features, unknown & unpredictable states, etc).
English
Steph Eaneff retweetledi
“I have learned to look when I want to look away. I have chosen to stay when I’d prefer to run out of the room and cry,” writes @SunitaPuriMD, a a palliative medicine physician. “The prelude to compassion is the willingness to see.” nyti.ms/3557a1r
English