Great piece on the important work @joan1k.bsky.social and co are doing
@bwaber
MIT Researcher, he/him, Senior Visiting Researcher @ Ritsumeikan, Co-Founder of Humanyze, former Senior Researcher @ HBS, author of People Analytics. AI, management, law, corporate governance, psychology, anthropology, ethics, and similar topics
Great piece on the important work @joan1k.bsky.social and co are doing
Still a good boy
Yep :-)
Oh yeah, that's like the next twenty slides. If I only have a minute, though, I normally start with that analogy
These are the most savage LinkedIn comments I think I've ever seen
Would not normally direct anyone to LinkedIn, but read the comments under the Grammarly CEOβs announcement about them pulling the βExpert Reviewβ feature. They are quite illuminating.
Yeah, they really need to understand interpolation at least to expand on that analogy. I'm going to stick with that analogy for now, imperfect as it may be
I'm super simplifying it for people with that analogy, which helps them understand it better than any other I've heard. I'm guessing you've read the paper by the Stanford folks showing that "emergence" is just a linear function of adding more data and not a step function
I disagree. They've filled in some pages with recipes from great chefs, but it tells you nothing about the other recipes
I would argue in terms of LLMs it's probably best thought of as random people as a base that is then fine tuned, but I think that the metaphor helps lay people better understand how they should think about these tools.
Best metaphor I've come up with for LLMs is a cookbook written by random people.
You wouldn't say a cookbook "knows" how to cook. Similarly, if a particular recipe was written by a great chef, it should be great. But if you don't know how to cook (or if you don't taste the dish) you can't tell
Last was "Fatherhood" by Augustine Sedgewick. This book is best viewed as a collection of short biographies of white men who made important contributions to Western thought over the millennia, with only some references to concepts of "fatherhood"
Full review: bookwyrm.social/user/bwaber/... (5/5)
Next was "Habits of Empire" by Walter Nugent. This book examines the long, continuous history of US territorial aggression, mostly vis-a-vis other Western powers. However. Native Americans are barely mentioned until chapter 8...
Full review: bookwyrm.social/user/bwaber/... (4/5)
Next was a fantastic discussion between @annmlipton.bsky.social & @activistinvestor.bsky.social on the EngageSmart controller lawsuit & a refreshing examination of why companies are constraining proxy advisors on the Shareholder Primacy Podcast. Highly recommend www.youtube.com/watch?v=9CJk... (3/5)
First was a great conversation with Robert Miller on liability-management exercises on the @busscholarship.bsky.social www.youtube.com/watch?v=bbGu... (2/5)
A selfie of me on a bridge above a wide part of the Charles River on a sunny day. Bare forest lines both banks, and some snow is still visible. I'm a bald, middle-aged, white man with a red beard flecked with white. I'm wearing black sunglasses and a dark blue running shirt.
It wasn't quite as warm as yesterday, but I still went out for a nice run and listened to talks for my #AcademicRunPlaylist! (1/5)
Announcement of the premiere of LINE, specifically touting the inability of loved ones to communicate during the East Japan Earthquake and tsunami as a selling point
A post on Twitter reminds people of something many have forgotten: When the Great East Japan Earthquake happened 15 years ago, LINE - now Japan's most popular messaging app with 100M users - didn't exist. It was actually developed in response to the disaster.
FWIW the lawsuit that was just filed is a class action, and if certified would require Grammarly to list all "experts" they used a la the Anthropic book lawsuit
I do think they've been overhauling a bunch of stations/tracks, but I'd obviously like to see some data on delays
Their service is so good, totally reasonable IMO
fuck yeah julia _get em_
Also we know a lot more about how the world works than we did in the 1800s
I would like to congratulate Grammarlyβs lawyers for belatedly realizing their shitty AI product was gonna get them class-action yeeted into the sun.
I would also like to encourage Grammerly to eat shit forever and ever amen.
The standard MO of physics, economics, and computer science π
Over the past 20 years, the raw number of criminal incidents involving foreigners has dropped by approximately 40 percent. Over the same period, the total foreign population in Japan nearly doubled, from 2.01 million to 3.95 million.
buff.ly/TkD9hkD
Last was a discussion on gender bias in AI systems at the LSE with @drpragyaagarwal.bsky.social, Karen Blake, Carrie Friese, and @lizstokoe.bsky.social (π) www.youtube.com/watch?v=1Aoa... (8/8)
Next was a thought-provoking session on AI's (non) impact on the labor market at @hamiltonproject.org with @jedkolko.bsky.social, Bharat Chandar, @marthagimbel.bsky.social (who knocks every question out of the park IMO), and Nathan Goldschlag www.youtube.com/watch?v=yiMn... (7/8)
Next was an excellent talk by Bedoor AlShebli on quantifying inequities across the scientific production pipeline at @iasliu.bsky.social www.youtube.com/watch?v=YiAu... (6/8)
Next was a compelling panel on opportunities and challenges in AI regulation at @hamiltonproject.org with @tarasinc.bsky.social, Catherine Tucker, and Nicol Turner Lee www.youtube.com/watch?v=TP7Q... (5/8)
Next was an engaging session on the present and future of industrialization at @voxdev.bsky.social with Tristan Reed, Rodimiro Rodrigo, Gustavo de Souza, and Markus Poschke www.youtube.com/watch?v=rX2d... (4/8)