eegilbert

However, the paper does offer a compelling causal mechanism. Also, it appears they performed follow-up pilot studies on the effects of erythritol consumption—as opposed to its simple presence in the blood. The following quote from the article is repeated in the linked Science commentary:

"The present studies suggest that following ingestion of an artificially sweetened food harboring typical levels of erythritol as artificial sweetener, plasma levels of erythritol remain elevated for many days, well above the thresholds necessary to enhance stimulus-dependent platelet reactivity, even among healthy volunteers."

I agree with Derek Lowe. I'll be avoiding erythritol until subsequent studies prove that it's safe.

Some of my immediate colleagues are hostile to this!

Agree with the substance of your point: often the importance of classes varies with the structure of the program (e.g., quals).

I need to fix that. Apologies. Somebody already caught it an earlier round of feedback, and I just haven't gotten to it yet with the pandemic. But will soon.

This is my syllabus. :-)

I tell all my prospective students to talk to my former students.

That's You Need A Budget, which I love paying for.

Page and Brin were both initially funded by the National Science Foundation to do the algorithmic work that led to Google [1]. Page was supported by an NSF grant from the 1994 Digital Library Initiative program; Brin was at Stanford on an NSF Graduate Student Fellowship.

[1] https://www.nsf.gov/discoveries/disc_summ.jsp?cntn_id=100660

One of the authors here. That's correct. The spillover to other sites is not studied.

Based on an (admittedly imperfectly) related paper with Stevie Chancellor & Munmun De Choudhury [1], I would expect that the spillover to other sites results in a 1) smaller, but 2) more committed group of users.

http://comp.social.gatech.edu/papers/cscw16.thyghgapp.chance... (see Summary of Findings, page 10)

Tanushree Mitra wrote a computational paper on this in 2016 [1], finding something similar:

"Using four years of longitudinal data capturing vaccine discussions on Twitter, we identify users who persistently hold pro- and anti-attitudes, and those who newly adopt anti-attitudes towards vaccination. After gathering each user's entire Twitter timeline, totaling over 3 million tweets, we explore differences in the individual narratives across the user cohorts. We find that those with long-term anti-vaccination attitudes manifest conspiratorial thinking, mistrust in government, and are resolute and in-group focused in language."

[1] http://comp.social.gatech.edu/papers/icwsm16.vaccine.mitra.p...

I agree with most of this. However, I might take issue with the motivation you ascribe to professors: "only car[ing] about increasing the number of academics who studied in their lab." Another explanation, one that I find myself feeling sometimes, is simply mismatched incentives. Practicing academics and students headed to industry simply have different incentives, and that can add friction to the relationship.

I think it would be helpful and useful for departments to consider other ways to facilitate industry pathways that don't so heavily rely on the advisor.

Here is the original SIGGRAPH paper and demo video:

http://grail.cs.washington.edu/projects/AudioToObama/siggrap...

https://www.youtube.com/watch?v=9Yq67CjDqvw

Perhaps. Although you could apply something like matching techniques [1] to find startups comparable to the big successes.

[1] https://en.wikipedia.org/wiki/Matching_(statistics)

I loved Founders at Work. However, it suffers from (somewhat severe) survivorship bias [1]. The next time around, it would be great to hear from more companies/founders that were "supposed to make it" but didn't for one reason or another.

[1] https://en.wikipedia.org/wiki/Survivorship_bias

It's really quite sad that someone with this much skill is spending this much time on a project like this.

Yes. This is often, perhaps willfully given the incentives, overlooked by computer scientists (and people trained in that discipline). For a complete, cogent argument see:

http://static.googleusercontent.com/media/research.google.co...

One example is this recent paper on learning (somewhat) high-level attributes of text from character streams alone (i.e., without telling the convolutional networks that things like words and punctuation exist).

https://arxiv.org/abs/1502.01710

Much better.

Yesterday, a startup incubator wants to be a research lab. Today, a government research lab wants to be a startup incubator.

I'm confused.

This looks great, as I would expect.

However, a major hurdle at US universities is FERPA [1]. FERPA (but really the fear of litigation under FERPA) is one reason why universities host their own software.

[1] https://en.wikipedia.org/wiki/Family_Educational_Rights_and_...

Hey look, it's my paper. :-)

Dunno why this caught back on now. Paper is from 2012.

I just read the Methods section of that article. It does not use a RCT design; it's an observational study based on questionnaire data. All standard confound caveats apply (with the exception of the controls accounted for in the paper).

Twitter lets your bot post thousands of times a day without then banning the account? Via the API?

I share these worries.

In addition, consider some of the potential social effects of flu season. You're on a crowded international flight and a handful of people are clearly ill. What are the odds someone on that plane freaks out about Ebola?

Reminiscent of folklore surrounding Airbnb's Craigslist campaign: http://davegooden.com/2011/05/how-airbnb-became-a-billion-do...

I think we should revisit the meaning of the word "democratic."

I sometimes use a personal OpenVPN install I have on AWS. I know from experience that Stack Overflow and Yelp block me; presumably, they block the whole range.