Dr. Sarah M Brown
@BrownSarahM
Followers
1K
Following
16K
Media
176
Statuses
10K
Assistant Professor of Computer Science | @alumniNU | @nsbe Lifetime Member | @theCarpentries instructor & trainer
Kingston, RI
Joined December 2011
We love our instructors! @BrownSarahM is teaching a @thecarpentries style Python workshop at #NSBE49 💙
0
3
17
Most brain-imaging studies make 3 questionable assumptions: mental events are localizable, map uniquely to dedicated #brain circuitry, & are independent of larger context. These 19th-century views need an update. New #OpenAccess paper in @TrendsCognSci. 1/
32
370
1K
If legacy admits are deemed Constitutional at universities, then admits based on being a descendant of slavery should be as well. Especially at universities such as UNC and Harvard that were financed by slave money and that prohibited descendants from enrolling.
233
2K
9K
The National Labor Relations Board—the sole federal agency tasked with enforcing private-sector employees' rights to organize and collectively bargain—is in a crisis. After years of inattention by Congress, the agency no longer has the resources to adequately enforce its mission.
23
1K
5K
Talking about "decolonial XXX" in the diaspora while supporting fascism at home, the cognitive dissonance.
131
273
642
Then why did you invite him to give a lecture this thursday??? 🐍🐍🐍 Entire fields of AI experts and you choose a mysognist to amplify? Also his book is obsolete, "Master Algorithm" was always cringe, and he hasn't published anything relevant in years
cs.washington.edu
#UWAllen leadership is aware of recent “discussions” involving Pedro Domingos, a professor emeritus (retired) in our school. We do not condone a member of our community engaging in a Twitter flame war belittling individuals and downplaying valid concerns over ethics in AI. 1/11
1
18
106
so, better here is.... more exploitive?
A computer is better at pricing units because, as one developer said, leasing agents have “too much empathy.” RealPage discourages bargaining with renters. 7/
0
0
0
we need to shift from talking about how amazing the current state-of-the-art generative models are to how these models steal from artists and what mechanisms for accountability should be put in place
Greg Rutkowski (@GrzegorzRutko14) is an artist with a distinctive style, known for creating fantasy scenes of dragons and epic battles. Rutkowski has now become one of the most popular names in AI art, despite never having used the technology himself. ⬇️ https://t.co/bxwWLK0Shv
0
42
132
This, for me, increase how much of a hassle it is to be a meta-reviewer/area chair by a lot. The actual work is not so bad, but the tools are always in the way.
0
0
0
CMT emails are almost always useless. it includes links to download CMT mobile apps (never going to happen) but not a link to the particular part of CMT where I could reply to the message or even a link to CMT at all.
1
0
0
In my other class, I taught number systems as a cultural artifact. This week was *all* socially embedded CS.
This week in my data science class, I introduced machine learning, but I did it in a very new way this year. In order to center thinking about the impact of algorithms we produce, I gave a high level overview of what ML is and then focused on model evaluation.
0
0
1
Anyway, here are the notes from this week. Part 1: https://t.co/a1rLR6OeEQ and 2:
0
0
8
After that, then I'll have them train their own models. I'm also hypothesizing that evaluating models that they did not train themselves will make it easier to be more critical and hoping that comfort with being critical carries over when they create the own.
1
1
10
It's early, but I think that students took the idea of multiple performance metrics more seriously this way. Their next assignment is going to be to audit a (simple) classifier, probably toy ones off of Adult (fair from aif360+ sklearn baseline) to practice interpretting metrics
1
0
3
In order to study evaluation *without* fitting any models, we used the @propublica COMPAS audit as an example. Our first interaction with ML was to *audit* the output of one, not to build one.
1
0
11
The framing being, you cannot do something well if you don't know how to tell if it works. Also, I think this an approachable way to integrate ethical thinking and ethical practice of data science.
1
0
12
This week in my data science class, I introduced machine learning, but I did it in a very new way this year. In order to center thinking about the impact of algorithms we produce, I gave a high level overview of what ML is and then focused on model evaluation.
2
4
43