Sign in

Michael Ekstrand

@md.ekstrandom.net
2.3K followers 1K following 3.3K posts

Info sci prof @ Drexel, trying to keep the machines (esp. RecSys & IR) from learning bigotry and discrimination. ADHDS9. Usually self-propelled. Opinions those of the Vulcan Science Academy. 🐰x2. 🏡 md.ekstrandom.net 🧪 inertial.science

PostsRepliesMedia
Michael Ekstrand @md.ekstrandom.net · 4h
Joe Konstan introducing the John T. Riedl Lasting Impact Award at #RecSys2026, recognizing work with a durable impact on the field. Happy to be a part of the committee who selected the first recipient and to carry forward John’s memory and legacy.
A photo of a convention room, with Joe on the stage and a slide with a brief bio summary of John on the slide.
021
Reposted by Michael Ekstrand
Bree (half of Kit Rocha) @mostlybree.kitrocha.com · 13h
your average Bluesky adventuring party has: - 1 sports person who is also a political sicko - 1 adhd chaos crafter who is also a political sicko - 1 sff writer who is also a political sicko - 1 romance author who is also a lawyer who is also a political sicko - 1 gardener recovering from being a po
2045474
Reposted by Michael Ekstrand
Ben Collins @bencollins.bsky.social · 22h
Gobsmacked by the elite levels of not knowing ball in the Senate today, thinking an NBA team is called the Hawkeyes then, refusing to Google it, then accusing a guy of treason because you didn’t know the Iowa Lady Hawkeyes aren’t in the NBA. Somehow I don’t trust these people to Save Girls Sports.
270126462081
Reposted by Michael Ekstrand
Maria Antoniak @mariaa.bsky.social · 29/09/2026
I'm recruiting 1-2 PhD students to join our lab in Fall 2027, through either Computer Science or Information Science at the University of Colorado Boulder! Looking for people with interests in NLP plus [healthcare | literary studies | narratives | social media | etc.]. Join us! 🏔️☀️
cls-lab.com
CLS Lab — Culture, Language, & Systems | University of Colorado Boulder
The Culture, Language, & Systems Lab at CU Boulder studies the language systems that transmit and shape modern culture, from internet platforms to literary archives to language models.
06750
Michael Ekstrand @md.ekstrandom.net · 29/09/2026
Love this concluding highlight from @mmitchell.bsky.social’s keynote at #RecSys2026: “let’s build systems that leave people better able to see, think, and say no”.
061
Michael Ekstrand @md.ekstrandom.net · 29/09/2026
Welcome to #RecSys2026!
020
Reposted by Michael Ekstrand
Kevin M. Kruse @kevinmkruse.bsky.social · 29/09/2026
When they get back in power, Democrats need to go all-in on higher education. We need a campaign of reinvestment in public universities across the country that makes the Morrill Land-Grant Act look like nothing by comparison *and* a GI Bill of Rights-style program of tuition grants.
20452221414
Michael Ekstrand @md.ekstrandom.net · 28/09/2026
Don’t know what you’re doing at #recsys2026 this morning? Come check out my tutorial on reproducibility! md.ekstrandom.net/pubs/recsys-...
md.ekstrandom.net
Continuous Automated Reproducibility
Upcoming RecSys tutorial on practical tools and workflows for reproducible recommender systems experiments.
041
Michael Ekstrand @md.ekstrandom.net · 28/09/2026
is this what it looks like to roll a 28 to troll the mathematicians?
020
Reposted by Michael Ekstrand
Michael "Shapes Dude" Betancourt @betanalpha.bsky.social · 28/09/2026
The logit function quietly smiles…
0123
Reposted by Michael Ekstrand
Will Quinn @willquinnart.bsky.social · 28/09/2026
Daily bunny no.3455 prefers the classics
Two bunnies are standing near a record player, and one starts to play their preferred record, "Something Else" by the Kinks. Next to him, the second bunny is miffed that she can't play her preferred record, "Room On Fire" by the Strokes.
132716526
Reposted by Michael Ekstrand
Brad Miller @bnmnetp.bsky.social · 26/09/2026
Such a great day at the grouplens 30 year reunion at the University of Minnesota. Great to reconnect with so many smart people.
011
Reposted by Michael Ekstrand
James Ryan @jdryan08.bsky.social · 26/09/2026
The combined undergraduate and graduate student populations of all Ivy League schools is about 140,000 students. The CUNY system alone is 250,000.
151359304
Michael Ekstrand @md.ekstrandom.net · 26/09/2026
Hello, Keller, my old friend…
A selfie of me in a black fleece shirt with Keller Hall, the UMN CS building, in the background.
000
Michael Ekstrand @md.ekstrandom.net · 26/09/2026
Arnold speaks here with greater conviction and moral clarity than a lot of pastors I've heard.
010
Michael Ekstrand @md.ekstrandom.net · 25/09/2026
Hello, Minneapolis! It’s good to be “home”.
011
Michael Ekstrand @md.ekstrandom.net · 25/09/2026
It’s an intense but fun start to the new academic year. Kamelot + Visions of Atlantis + Frozen Crown in Glenside last night, leaving for GroupLens reunion and RecSys this afternoon.
010
Michael Ekstrand @md.ekstrandom.net · 25/09/2026
tfw you roll a 38 to hit 😳, but forget your class feature that should have made it a crit 😢.
100
Michael Ekstrand @md.ekstrandom.net · 24/09/2026
gratitude of the day/week: rabbit’s skin thingy was not cancerous. picture of rabbit with her octopus (by @bugsnbuns.bsky.social) for pet tax.
A black lionhead rabbit loafing by a red plush octopus.
030
Reposted by Michael Ekstrand
Faine Greenwood @faineg.com · 24/09/2026
It really can’t be stressed enough that the freaky AI doom cult stuff that is so influential in Silicon Valley is an AMERICAN phenomenon. The AI industry in Europe and in Asia is not, from all I’ve heard, taking this weird millenarian stuff seriously. But much of US Big AI? It very much is.
2239498
Michael Ekstrand @md.ekstrandom.net · 24/09/2026
Is it normal for arXiv submissions to be "on hold" for nearly a week now? Our forthcoming RecSys Research & Practice Note has been stuck in the queue since last week 😢.
110
Reposted by Michael Ekstrand
Alex Kale @kalealex.bsky.social · 24/09/2026
What AI scales about science are the parts that are most broken and gameable. With the exception of combinatorial problems like protein folding, we need more thinking and less rote computation.
0236
Reposted by Michael Ekstrand
jamelle @jamellebouie.net · 23/09/2026
i think you can fairly say that the second trump administration is the most racist federal administration since woodrow wilson and honestly i'm not sure that it isn't more racist than wilson's
164113112103
Michael Ekstrand @md.ekstrandom.net · 23/09/2026
I’m a pretty good programmer (for certain technical/elegance-focused definitions of good, at least). Every damn time I point a competent model at some of my code, it finds meaningful ways to improve it, consistency issues I missed, etc. I don’t know how I feel about this.
130
Reposted by Michael Ekstrand
Carlos Scheidegger @cscheid.net · 23/09/2026
Louder for the people in the back: Jev can’t be calibrated www.alexmolas.com/2026/09/23/j... Friends don’t let friends lie about the accessibility of the DGD even in principle
alexmolas.com
Jev can't be calibrated
Jev is a useful zero-shot classifier, but its probabilities can't be calibrated for your data. Calibration depends on your data distribution, which Jev never sees, so treat its outputs as scores and r...
44410
Michael Ekstrand @md.ekstrandom.net · 23/09/2026
What the everlasting nope? www.reddit.com/r/sysadmin/c...
reddit.com
From the sysadmin community on Reddit: PSA: Grammarly will send unhinged messages to all your users if you try to cancel
Explore this post and more from the sysadmin community
010
Reposted by Michael Ekstrand
Micah @rincewind.run · 22/09/2026
stand up for trans people: it is the right thing to do and it just will not hurt you, not now, not in this environment and maybe folks can start to relearn that trans people are pretty fucking cool, actually
51133174
Michael Ekstrand @md.ekstrandom.net · 22/09/2026
shit like this inclines me to believe in a multiverse, because a universe in which these clowns (Bret, not Maris) make a living for ideas cannot be the best universe in existence. i mostly jest, but still…
010
Reposted by Michael Ekstrand
Drew Starr @thedrewstarr.bsky.social · 22/09/2026
"is a hot dog a.sandwich" was a load-bearing non-debate debate that got unfairly stigmatized as pointless ragebait when it was actually an important release valve
0153
Michael Ekstrand @md.ekstrandom.net · 22/09/2026
Shall We Find Something to Kill? Personalized Bounty Recommendation for Droids and Meatbags
static.klipy.com
HK-47: Star Wars KOTOR Droid with Rifle
Alt: HK-47 from KOTOR, raising his rifle.
010
Reposted by Michael Ekstrand
Pookleblinky @pookleblinky.bsky.social · 28/02/2026
That is *all* a chatbot offers: unaccountability. A drone-strike on a nursery being just as unaccountable as a waymo running a red light, with just as little consideration at all given to it legally or ethically.
139980
Reposted by Michael Ekstrand
Robin Burke @robin-burke.bsky.social · 22/09/2026
On Thursday, I get to relax while @md.ekstrandom.net presents our TORS highlight paper🏅and Solé Pera, @julianeidhardt.bsky.social, @navatintarev.bsky.social and Özlem Özgöbek present our Women in RecSys winning paper. 🏆 6/8
111
Reposted by Michael Ekstrand
Robin Burke @robin-burke.bsky.social · 22/09/2026
A busy week for That Recommender Systems Lab at RecSys 2026! WHERE TO SEE US: We start off the week with Elizabeth McKinnie and @aishwarya-satwani.bsky.social at the Doctoral Symposium (not public but you can catch their posters on Monday). 1/8
131
Michael Ekstrand @md.ekstrandom.net · 21/09/2026
me: needs to write lecture slides. also me: drafts Spice Girls paper titles.
050
Michael Ekstrand @md.ekstrandom.net · 21/09/2026
Leaving at the end of the week for #RecSys2026, where my students & I are doing a lot of cool stuff. Read more about it from Drexel SCIS, and hope to see you in Minneapolis! drexel.edu/cci/news/202...
drexel.edu
Strong Howley College Presence Planned for RecSys 2026 - School of Computer and Information Sciences
030
Michael Ekstrand @md.ekstrandom.net · 21/09/2026
Adding taint support to my HTTP client script code, and running into an interesting Q. I'm requiring URLs to be untainted, but assuming the HTTP client library properly escape values, should query parameters and headers also be taint-checked?
000
Reposted by Michael Ekstrand
Imani Gandy @angryblacklady.blacksky.app · 20/09/2026
I just honestly can't get over that Save Women Sports, an anti-trans group could only find five trans athletes competing K through 12. Five. The NCAA president testified a year and a half ago or so that there were less than 10 college level trans athletes. What the fuck are we doing, people
16886362295
Reposted by Michael Ekstrand
hyla @hylabuns.bsky.social · 21/09/2026
Fizgig says: be gay
My bunny Fizgig stands in some rainbow light
1025668
Reposted by Michael Ekstrand
Renee DiResta @noupside.bsky.social · 21/09/2026
The ceNsOrShiP iNdUstrIaL cOmpleX hoaxers are starting to go after AI evaluators using the same playbook that took down social media research. It worked for them last time. No consequences for lying. Shellenberger, Schmitt both starting up. Smears & subpoenas from the lie machine commencing. Watch.
Anthropic's CEO is pushing America's Al labs to embed a leftist alignment apparatus.
But he's deeply intertwined with this ecosystem, its funding, its people, and the supposed experts shaping the media coverage.
2
178
89
Ill 2.5K
Eric Sch...
@Eric_Sch... •7h
I want the grant records.
The stock transfers.
The board ties.
The conflict policies.
The journalism disclosures.
•••
They will all be hearing from me soon.
No self-appointed leftist cartel gets to decide America's future.
48427
Michael Ekstrand @md.ekstrandom.net · 21/09/2026
Prepping for the first week of the term with a fresh notebook, Hakushu single-malt whisky, and Ukrainian progressive metalcore.
A white album cover for Duel by Jinjer on top of a black turntable, with a burgundy notebook and a light amber whisky.
140
Reposted by Michael Ekstrand
Anthony Moser @anthonymoser.com · 20/09/2026
when the fascists say we just need to get rid of people, if your response is to argue about which people and how, you have already stepped into their frame to play their game of make believe you need to eat their frame: nobody goes!
335275
Reposted by Michael Ekstrand
Anthony Moser @anthonymoser.com · 20/09/2026
fascists only have one move, which is to find people to blame problems on. they can't address any real problems bc that's not part of the story. they just have to find people to blame. and anybody who doesn't play along becomes somebody to blame
237248
Michael Ekstrand @md.ekstrandom.net · 20/09/2026
my silly Tcl project has basic HTTP client support now… tcl.mde.one/guardian/htt...
tcl.mde.one
HTTP support - Jim Tcl in Rust
100
Reposted by Michael Ekstrand
value plus discounted @akhilrao.bsky.social · 19/09/2026
economists metascientists 🤝 "gold standard is not good"
13313
Reposted by Michael Ekstrand
Dr. Emily Friedman @ecfriedman.com · 19/09/2026
Several folks are rolling “crowd sourced” TTRPG-related databases. I love data! But I love SUSTAINABLE projects. Any site aiming to collect data (ESPECIALLY now) should have a page detailing: - metadata structure - initial seed corpus - intended audience - tech documentation - interoperability
413031
Reposted by Michael Ekstrand
Adam Becker @adambecker.bsky.social · 19/09/2026
@aoc.bsky.social just put out a great statement on AI. Refreshing to see a major politician on the left nail this. "Silicon Valley billionaires want us to believe that Al robots will destroy humanity. But the biggest threat to humanity are the tech oligarchs themselves." 🎯
Charles,

Last weekend, Dario Amodei, the CEO of Anthropic, one of the largest Al companies in the world, publicly asked his industry to slow down Al development.

Shortly after, Sam Altman, the CEO of OpenAl, followed suit. Then Elon Musk.

This all came days after an Anthropic researcher quit, citing his belief that this technology could wipe out humanity and that Al companies are "gambling with our lives."

For years, people have warned that generative Al could slip out of our direct control.

Those safety concerns must be taken seriously.

And this is why I have advocated for strict safety standards on how Al is built and deployed, public oversight over this technology, and using existing laws to hold these companies accountable when they cause harm.

But I want to draw your attention to another reality: Right now, the U.S. economy and our financial futures are riding on Silicon Valley's bet that Al will soon be immensely profitable and pay off. And that bet isn't looking good.

The formula for a profitable business isn't complicated: make more money than you spend. But the biggest Al companies aren't doing that. They are spending hundreds of billions per year on data centers, chips, and other costs while their ability to make revenues that exceed their already-immense sunk costs are unclear. Experts estimate that these Al companies will require another $600 billion of debt just to financethemselves in 2026 alone. That is more than the entire budgets of the U.S. Departments of Justice, Transportation, & Education combined.
Meanwhile, despite almost every single one of these companies not turning a profit & having sunk billions of investor dollars in development of data centers & other infrastructure, Al companies are also on a spending spree of hundreds of millions of dollars in lobbying & dark money in our elections.
It would be one thing if we were discussing a few private companies with a poor business model. But last year, virtually all stock market gains - 80% - rested entirely on just a handful of Al companies.
That tells us two things:
1. Nearly the entire U.S. economy outside the Al industry is relatively stagnating and faring poorly & the gains from these few companies are disguising it; &
2. Millions of people's retirements - including their pensions, 401ks, & investments - are potentially exposed to this risk.
That's why these Al companies' dubious financial positions matter to working people.
These companies argue that debt is necessary. They'll tell you that building transformative technologies requires upfront capital, & that massive borrowing today will pay off in generating unprecedented prosperity tomorrow.
But this is not the typical debt that startups take on while they build out a company. We are talking about billions going on trillions of dollars leveraged with no indication that these companies will be able to recoup these funds profitably. And they are increasingly looking to the government as their source of cash.
In an effort to hide their debt, tech companies have been relying on "circular financing" where tech giants like Google, Amazon, & Microsoft give billions to Al startups like OpenAl & Anthropic, only for those startups to hand that money right back to buy Al processing power.
In other words, these Al companies are luring in new investors by making it look like their financials are healthier than they really are.What's even worse is that retirement & pension funds are key to supporting this illusion.
When the Al companies issue bonds to raise more money, your pension funds & retirement accounts are some of the biggest buyers. When Musk's Al business went public, it became part of index funds where millions of Americans invest our savings. Many Americans don't even know that their retirement investments are tied up in these companies at all.
From Big Tech's perspective, that's a feature, not a bug. By tying our futures to their reckless gambling, Al billionaires are trying to make themselves too big to fail. They think they can take such dangerous risks because we will bail them out if things go wrong.
Sound familiar? This scene is dangerously reminiscent of the rampant speculation, corruption, & broken incentives leading up to the 2008 financial crisis that plunged millions of working class Americans into financial chaos that we never fully recovered from.
It is no accident that these warnings from Altman & Amodei come months-if not weeks-before these companies were meant to go public.
When companies go public, they have to be more transparent about how they are doing financially. And after a waiting period, they typically have to let private investors-like their own employees-sell their shares.
The New York Times recently confirmed that OpenAl has postponed going public for this very reason. They are in no rush for the truth about their finances to come out.
But the moment these oligarchs began to seek our public tax dollars & public contracts while exposing their companies to public pension funds & investments, they opened themselves to public accountability & scrutiny.
Silicon Valley billionaires have rigged the game & are attempting to shift their financial risk to working people: If they're right about Al's potential, they reap the profits. If they're wrong & the bubble bursts, they are setting up the American people to pick up the tab.
Heads, they win; tails, we lose.We can stop this, and we shouldn't play their game. Together we can address both the economic and safety threats of unrestricted Al development and prevent financial chaos and stabilize market risk.

We must make it clear now that should the Al bubble pop, Congress should not bail out Big Tech.

And across our federal agencies, we have the authorities and tools to hold Al companies and their leaders accountable for releasing dangerous, unvetted, and defective products. We must not wait.

Silicon Valley billionaires want us to believe that Al robots will destroy humanity.

But the biggest threat to humanity are the tech oligarchs themselves.

It's time we hold them accountable.

Thank you for reading.

In solidarity,

Alexandria
221562596
Reposted by Michael Ekstrand
Nicholas Grossman @nicholasgrossman.bsky.social · 19/09/2026
We already have an AI kill switch: cut power to servers. Really want to talk existential threat? Fine. We humans have many ways to blow things up. Like data centers. Panicky discourse assumes the first priority is protect tech company investments and expand their businesses. But why should it be?
441632323
Reposted by Michael Ekstrand
Shannon Vallor @shannonvallor.bsky.social · 19/09/2026
This is going around as essential reading and for very good reason. Honestly the most incisive and full summary of the longtermism/x-risk pathology I have seen. Every stanza is a gem, and the closing is genius, but my personal favorite, because I have had this exact conversation almost verbatim:
[nitter.app post excerpt from banteg]

>someone asks where the one-in-a-million number came from
> expert judgement
>which expert
>us
>how calibrated
> extremely thoughtfully

›reduce estimate to one in ten million to be conservative
›still beats curing cancer by 38 orders of magnitude
›epistemic robustness achieved
87825
Reposted by Michael Ekstrand
Faine Greenwood @faineg.com · 18/09/2026
In isolation, I’m OK with futurists. I’m OK with futurists who have wild and crazy ideas about long term risks. The problem is that we somehow have let the kooky little “what if Dyson spheres had feelings” guys almost entirely take over the US AI regulatory space, and that’s *really bad*.
1230948