1. Bluesky Feeds /
  2. Andrew Heiss /
  3. Andrew Heiss

Feeds Stats

  • 💙 Liked by 0 users
  • 📅 Updated almost 2 years ago
  • ⚙️ Provider skyfeed.me

Andrew Heiss Likes over time

Like count prediction
The feed Andrew Heiss has not gained any likes in the last month.

Feed Preview for Andrew Heiss

Andrew Heiss
@andrew.heiss.phd
about 24 hours ago
keep thinking about how higher ed is out here enthusiastically embracing and pushing LLMs for everything and I despair
4
2
44
Andrew Heiss
@andrew.heiss.phd
1 day ago
once again revisiting @adamkotsko.bsky.social 's fantastic essay here itself.blog/2025/07/27/c...
I hate generative AI. I hate how it’s destroying writing pedagogy and giving students even more excuses not to read (because they can just read a “summary”). I hate how whiny and defensive AI users are about the pathetic little ways they’ve integrated it into their lives. If I could push a button and permanently delete it from existence, I would. If I could go back in time and prevent it from being invented, I would.

The reason I hate it is not just that its output is mediocre bullshit. It’s that it is an active attack on everything I value — literacy, analysis, thought. I find the
2
11
53
Andrew Heiss
@andrew.heiss.phd
1 day ago
grading final projects and I've never ever seen so much brazen LLM output. **Most stuff is still great!** only ≈20ish% is pure LLM (text like "Lifenpsctarece.years" in completely made up images!) This is the most disheartening time I've had grading, like, ever. What are we even doing anymore.

On the verge of declaring defeat with chatgpt in my asynchronous online dataviz class. Something changed this semester compared to past ones and SO MANY assignments are essentially 100% LLM output.

13
23
273
Andrew Heiss
@andrew.heiss.phd
3 days ago
5 of 5 stars to Eleanor Roosevelt, Vol 1: 1884-1933 by Blanche Wiesen Cook www.goodreads.com/book/sh….

www.goodreads.com

0
2
6
Andrew Heiss
@andrew.heiss.phd
5 days ago
ahhhhhhh our oldest is moving out today—she's off to BYU!
9
1
55
Andrew Heiss
@andrew.heiss.phd
6 days ago
Finally got to read this new paper by @cbarrie.bsky.social & @lexipalmer.bsky.social & Arthur Spirling on the lack of replicability in LLM-based research and polisci and it's so good and concise and well-reasoned! arthurspirling.org/docume….
These jobs range from coding the ideology of manifestos to identifying types of protest events to detecting certain political valences in speeches. The idea is to precisely calibrate exactly
how replicable one can expect machines to be in practice, and where (what types of tasks) we can expect better (lower variance, more replication) or worse performance. The human
workers provide a baseline comparison in terms of replicability. At a high level, the news is bad: while it is true that LMs can be (very) accurate relative to a gold standard, they also show considerable variance over time. And this is to say nothing of cases where they simply will not run at all, and thus fail the most basic requirement (see, e.g., Benureau and Rougier, 2018) of computational replication. Contrary to popular belief, the problems do not go away even if one sets "temperatures" (or equivalent tunings) to zero; indeed, this induces new but unpredictable problems with replication. Unsurprisingly, this variance affects the substantive answers we get downstream-that is, in subsequent analysis in which the labels
3.2 The Problem with Language Model Replication
The central problem with replication for Language Models is that as we will show-the process exhibits the weaknesses of deterministic, stochastic and rule-based replication, without the strengths of any of them. To make this point clear, consider Table 1. There we document replication practices as a typology. What defines the typology is first, whether exact replication is possible; second, whether replication is fragile in the sense we discussed above.

With a 2x2 table showing that LLms are not exactly replicable and are fragile
Full results for each outcome and run are displayed in Figures 8 to 10 in SI C. We also give descriptions of what we found. For now, we summarize our main observations:
1. For the manifestos, the crowdworkers perform very well (by LM standards) and their variance is generally lower than the LMs.
2. For the protests crowdworkers are less accurate than the LMs, but very consistent in their performance.
3. Crowdworkers struggle in predictable ways: for example, they are least accurate when manifestos should have 'extreme' codings (far left /far right).
4. LMs struggle in unpredictable ways: for example, GPT made errors on more moderate (liberal manifestos, but it is hard to know why.
5. Comparing across LMs, errors and performances appears to be idiosyncratic: for example, Llama has recall on some tasks on a par with GPT but generally much lower
variance.
6. Open LMs have the best replication performance, at least in terms of low variance.
For instance, on the static tasks, Llama has practically zero variance in its coding performance.
3. Consider open models that allow offline versioning. We found that, uniquely, our open-weights implementations were replicable to a high standard if that standard is low variance. That is, if the goal is something approaching the Deterministic 'code and data' replication vision above, then local, versioned models are the way to go. These may not deliver top of the line performance (e.g. accuracy) but should be checked as a first resort. We acknowledge that an open LM may not be "transparent" in the sense that it is "easy" to understand how it produces predictions even if one has the weights. But it is obviously a boon to replication insofar as being able to verify that
the original researcher did indeed see the results they reported. What is more, recent research into LM interpretability points the way toward more model understanding and control but only if weights are accessible (Cunningham et al., 2023).
2
10
37
Andrew Heiss
@andrew.heiss.phd
6 days ago
Checking in a kid at urgent care and have to choose a primary language and I’m amazed at the list they’re using. Almost chose English, old or Egyptian (ancient)
List of languages like

Dzongkha
Eastern frisian
Efik
Egyptian (ancient)
Ekajuk
Elamite
English
English, middle (1100-1500)
English, old (ca.450-1100)
Erzya
Esperanto
Estonian
Ewe
Ewondo
12
4
53
Andrew Heiss
@andrew.heiss.phd
8 days ago
been fighting with local DNS issues all day where adguard home stops working(?) and takes down the whole internet at home, ooof
1
0
3
Andrew Heiss
@andrew.heiss.phd
8 days ago
ahhhh this @acollierastro.bsky.social takedown of vibe physics and LLM stuff in general is phenomenal
vibe physics

www.youtube.com

vibe physics

YouTube video by Angela Collier

2
3
20
Andrew Heiss
@andrew.heiss.phd
10 days ago
pizza night with three different crust types (regular, sourdough, sourdough + cornmeal) #poliscicooks #pacooks
5 pizzas (with two more off camera in the oven)
2
2
46
Andrew Heiss
@andrew.heiss.phd
11 days ago
just like buying a new empty notebook, switching todo apps will *definitely* make me more productive
5
6
90
Andrew Heiss
@andrew.heiss.phd
12 days ago
We found a Pelican’s! We haven’t had this since we lived in Durham!
Kid standing next to a Pelican’s sign
1
0
4
Andrew Heiss
@andrew.heiss.phd
12 days ago
A bunch of the Bullochs worked/fought for the Confederacy and weren’t allowed back in the country after the civil war, which seems like a wise move

I’m reading a biography of Eleanor Roosevelt and it mentioned that Teddy’s mom grew up in Roswell, Georgia, which is like 15 minutes from my house, so we went and found Bulloch Hall, which is still here and a museum now

Bulloch Hall
Sign in the dining room saying that Teddy Roosevelt’s parents were married there
Slave quarters
0
3
6
Andrew Heiss
@andrew.heiss.phd
12 days ago
I’m reading a biography of Eleanor Roosevelt and it mentioned that Teddy’s mom grew up in Roswell, Georgia, which is like 15 minutes from my house, so we went and found Bulloch Hall, which is still here and a museum now
Bulloch Hall
Sign in the dining room saying that Teddy Roosevelt’s parents were married there
Slave quarters
2
7
55
Andrew Heiss
@andrew.heiss.phd
14 days ago
these are so well made! (via @djinnandtonic.bsky.social at www.etsy.com/shop/WideAwa... )
Color wide awake Lincoln pin
Stainless steel wide awake pin
0
1
10
Andrew Heiss
@andrew.heiss.phd
14 days ago
New blog post! @posit.co Positron doesn't use Rproj files tofor #rstats projects—instead you open folders in the IDE directly. Opening folders can be a little tricky though. This shortcut lets you right click on a folder in macOS Finder to open a it in Positron! www.andrewheiss.com/blog/….
7
6
35