iiinova — where your digital self keeps evolving

Reposted by Stephen Wild

Check out this talk by Ludwig Schmidt from a few years ago. The transformer is important not because of some magical geometry fact, but because it's the most efficient at memorizing massive datasets given our current hardware constraints. simons.berkeley.edu/talks/ludwig...

simons.berkeley.edu

Are Transformers necessary? A data-centric view on generalization

Researchers have proposed many methods to make neural networks more reliable under distribution shift, yet there is still large room for improvement. Are better training algorithms or training data th...

292

Stephen Wild @stephenjwild.bsky.social · 12h

All right, I'll play. Science and engineering will continue to improve human welfare

1100

Stephen Wild @stephenjwild.bsky.social · 14h

Up next in my attempt to understand the appeal of John Wayne's manliness (Inspires by reading Jesus and John Wayne, and John Wayne's America)

Cover image for John Wayne’s World: Transnational Masculinity in the Fifties.

Author: Russell Meeuf

130

Stephen Wild @stephenjwild.bsky.social · 17h

Very confused as to why a bunch of you keep talking about Hawkeye from The Last of the Mohicans

210

Stephen Wild @stephenjwild.bsky.social · 18h

Were they not voluntarily managing it before?

130

Reposted by Stephen Wild

Sanjay Srivastava @sanjaysrivastava.com · 30/09/2026

Guy who thinks chi-square fit tests are liberatory because they want all parameters to be free

1153

Reposted by Stephen Wild

Thomas Lumley @tslumley.bsky.social · 29/09/2026

Optimal Regular Additive Non-biased Gaussian Estimator

031

Stephen Wild @stephenjwild.bsky.social · 29/09/2026

Guy who thinks linear regression is progressive because the outcome is on the left

3252

Stephen Wild @stephenjwild.bsky.social · 29/09/2026

To give a serious answer to Liam's question: This is an extreme case, which makes the dynamics easier to see. It's not necessarily for there to be a raw, visceral hatred of women. Many times it's that men care more about the approval of men than they care about the consequences for women.

1120

Stephen Wild @stephenjwild.bsky.social · 29/09/2026

THANK YOU FOR YOUR ATTENTION TO THIS MATTER

040

Reposted by Stephen Wild

Malcolm 朝精 Barrett @malcolmbarrett.malco.io · 29/09/2026

meanwhile Phase 1 Phase 2 Phase 3 LLMs ? AGI

062

Stephen Wild @stephenjwild.bsky.social · 28/09/2026

I wonder how these respondents interpet "too soft and feminine"

A chart showing the 80% percent of Republican men and 70% of Republican women say that society has gotten too soft and feminine, up from.about 70% and 40% respectively.

The numbers for Democratic men and women are 20% and 14%, down from 40% and 20%.

140

Stephen Wild @stephenjwild.bsky.social · 28/09/2026

A genuine question here is if we can teach crows to give naloxone to non-responsive ones. Crows are smart. I bet we can. www.cbc.ca/news/canada/...

Crows being given naloxone after suspected drug overdoses in Prince George, B. C.

110

Reposted by Stephen Wild

Ben Bolker @bbolker.bsky.social · 28/09/2026

Stats explorations w/ Claude remind me of "a 4-wheel drive vehicle just gets you stuck in deeper mud". Going down rabbit holes I would previously have avoided ... surprisingly (to me) small/complicated differences between LRT & Wald for coverage/type-1 error ... bbolker.github.io/bbmisc/wald_...

bbolker.github.io

Wald vs likelihood ratio tests: examples

3354

Stephen Wild @stephenjwild.bsky.social · 27/09/2026

existence of liberal sciences implies existence of skimpy sciences

030

Stephen Wild @stephenjwild.bsky.social · 27/09/2026

Learned this summer that dog is faster than bird. Bird survived, but traumatized.

141

Stephen Wild @stephenjwild.bsky.social · 27/09/2026

I mean, one of the things Thayer did at West Point was reduce humanities and emphasize math, engineering, etc

000

Stephen Wild @stephenjwild.bsky.social · 27/09/2026

Everybody already hates Mr. Beast

000

Reposted by Stephen Wild

Sanho Tree @sanho.bsky.social · 26/09/2026

Happy Stanislav Petrov day! He saved the world from nuclear annihilation on this day in 1983. My freshman year in college would have been so much worse without his courage.🧵

5599185

Stephen Wild @stephenjwild.bsky.social · 26/09/2026

Red sky at night, Sailors's delight...

130

Stephen Wild @stephenjwild.bsky.social · 25/09/2026

Pumpkin spice everything will do that to you

020

Stephen Wild @stephenjwild.bsky.social · 25/09/2026

Always fun when Quebec politics makes an appearance here

240

Stephen Wild @stephenjwild.bsky.social · 25/09/2026

There have been some moments where Canada didn't consider things to be great (eg, Fenian invasions, McKinley tarrifs, Diefenbaker and JFK, invasion of Iraq, etc)

150

Stephen Wild @stephenjwild.bsky.social · 25/09/2026

62 takes carefully edited into one long, continuous shot

020

Stephen Wild @stephenjwild.bsky.social · 25/09/2026

Lots of prestige TV, not so much mass market TV. what are the titles of the most popular shows send tweet.

460

Stephen Wild @stephenjwild.bsky.social · 25/09/2026

Statistical pragmatisy: I'll use whatever works, with a preference for Bayesian methods. Why? Bayesian methods came along at the right time for me, and they let me answer a question I could not using other tools at the time.

470

Stephen Wild @stephenjwild.bsky.social · 25/09/2026

This is correct. And also the entire discourse around bringing iced coffee to an interview is annoying.

020

Stephen Wild @stephenjwild.bsky.social · 25/09/2026

Star Trek II: The Wrath of Sal Khan

070

Stephen Wild @stephenjwild.bsky.social · 24/09/2026

Levitin's "The Organized Mind" has a section on cognitive offloading, and a bunch of it can be read as "be rich and hire an assistant"

171

Stephen Wild @stephenjwild.bsky.social · 24/09/2026

Dying from laughter: “Listen, I have a great guy to tell you about. He’s such a good match for you. Now hear me out: He’s in tech.”

050

Stephen Wild @stephenjwild.bsky.social · 24/09/2026

Ahhh, yes. Time for half of Bluesky to go, "These are the stupidest voters imaginable. You get what you voted for, blah blah" The other half goes, "These are typical voters. You need to go out and talk with people" And the remainder try to stir the pot by putting the two sides against each other.

050

Stephen Wild @stephenjwild.bsky.social · 24/09/2026

Nothing. That's what Twitter is for

0110