Alex Saum
LightNews LightNews
banner
alexsaum.bsky.social
Alex Saum
@alexsaum.bsky.social
1.1K followers 1.8K following 230 posts
reader, writer, poet, professor. Doing digital things in and out of UC Berkeley. En inglés y en español. Spain and LatAm, but also English comparative stuff <3 (she/her)

http://www.alexsaum.com/

// not sure if she can do this all over again //
Posts Replies Media Videos
Pinned
alexsaum.bsky.social
Alex Saum @alexsaum.bsky.social · Oct 22
Stop scrolling and take 5 minutes to read this poem I published in Los Angeles Review. Stop resisting and just do it.

losangelesreview.org/resistance-t...

Ps: I promise there's no virus hiding in the zip file. Maybe. Stop resisting.
Resistance to by Alex Saum - The Los Angeles Review
Where is the poem? I google the obvious and the algorithm spits back Poe’s “TheRaven” to me. I find this surprising, but not completely wrong. I ask an AI about this search result, and the bot respond...
losangelesreview.org
Reposted by Alex Saum
hypervisible.blacksky.app
Today in luxury surveillance.
parismarx.com Paris Marx @parismarx.com · 2d
There are many more bad tech products to avoid. But I keep coming back to @hypervisible.blacksky.app and David Golumbia’s concept of luxury surveillance.

We need to really think about the surveillance tech we’re adopting — and certainly that we’re pushing on others.

disconnect.blog/a-tech-criti...
Ultimately, a lot of these products make me think of a concept called luxury surveillance coined by Chris Gilliard and the late David Golumbia a few years ago. They were arguing that there’s a line where forms of surveillance can feel oppressive to some — particularly poorer and marginalized groups in society — but that more privileged groups can feel are actually benefiting from them by making their lives easier or more convenient. Yet, Gilliard and Golumbia argue that line can easily shift, and while the surveillance may once have felt it was worth the trade off, it can quickly become an oppressive force for a much wider swath of people. “We need to develop a much deeper way of talking about surveillance technology and a much richer set of measures with which to regulate their use,” they wrote. “Just as much, we need to recognize that voluntarily adopting surveillance isn’t an isolated choice we make only for ourselves but one that impacts others in a variety of ways we may not recognize.” That’s something to consider in our everyday lives as much as when we’re choosing what to gift to others.
November 25, 2025 at 5:35 PM
alexsaum.bsky.social
And maybe one day we may even realize that "language" and "text" or even "writing" are not the same thing either; and definitely not synonymous of "intelligence"
reckless.bsky.social nilay patel @reckless.bsky.social · 2d
I’ve been running around asking tech execs and academics if language was the same as intelligence for over a year now - and, well, it isn’t. @benjaminjriley.bsky.social explains how the bubble is built on ignoring cutting-edge research into the science of thought www.theverge.com/ai-artificia...
November 25, 2025 at 5:34 PM
alexsaum.bsky.social
👀 Friends! The CPF for the next ELO conference is out!! 🚀

ELO26 (UN)SUPERVISED will be ONLINE only 🌐

July 15-18 & accessible from anywhere from your computer <3

CFP here>>>> projects.cah.ucf.edu/elo2026/cfp....

#elit #digitalliterature #netart #elo @eliterature.bsky.social
Logo display of ELO26 (UN)SUPERVISED. ONLINE. Accessible anywhere where from your computer July 15-18, 2026 - CFP out now!
November 25, 2025 at 5:03 PM
Reposted by Alex Saum
courtneymilan.com
I think it’s pretty clear at this point that one of the main impacts of LLMs is to disrupt thinking: to make it so that far too many people never properly learn how to do it, and then to control the output so there are thoughts that people never learn how to think.
November 24, 2025 at 5:18 PM
alexsaum.bsky.social
More languages and less blueprint projects, thanks
natureportfolio.nature.com Nature Portfolio @natureportfolio.nature.com · 3d
Multilingualism is associated with a reduced risk of accelerated ageing, according to data from more than 86,000 people across 27 European countries published in Nature Aging. go.nature.com/4hWRpLh #medsky 🧪
This is figure 2, which shows BAG results and epidemiological associations.
November 24, 2025 at 2:54 AM
Reposted by Alex Saum
natureportfolio.nature.com
Multilingualism is associated with a reduced risk of accelerated ageing, according to data from more than 86,000 people across 27 European countries published in Nature Aging. go.nature.com/4hWRpLh #medsky 🧪
This is figure 2, which shows BAG results and epidemiological associations.
November 24, 2025 at 2:44 AM
Reposted by Alex Saum
rbtownsend.bsky.social
"While other universities report that the humanities are shrinking, at Berkeley, the opposite is true. The music major is the fastest-growing major on campus. We are finding bigger classrooms because film is exploding. English is back to the numbers we saw 15 years ago. We are hiring" bit.ly/4ohKuOe
"The humanities really are a resource — a confidence for living in our times.” Dean Sara Guyer on the modern utility of humanities degrees
This interview originally appeared on the Division of Arts
bit.ly
November 23, 2025 at 3:38 PM
alexsaum.bsky.social
“These findings demonstrate that stylistic variation alone can circumvent contemporary safety mechanisms, suggesting fundamental limitations in current alignment methods and evaluation protocols.”

If this doesn’t sound like what a love poem generally does, I don’t know what does
alexsaum.bsky.social Alex Saum @alexsaum.bsky.social · 6d
I said it before and will say it again: poetry will break us, poetry will heal us and apparently now it will also… crack LLMs? Adversarial Poetry as a Universal Single-Turn Jailbreak Mechanism”
cstross.bsky.social Charlie Stross @cstross.bsky.social · 7d
Looks like LLMs are *very* vulnerable to attack via poetic allusion: "curated poetic prompts yielded high attack-success rates (ASR), with some providers exceeding 90% ..."

https://arxiv.org/html/2511.15304v1
November 21, 2025 at 3:01 AM
alexsaum.bsky.social
I said it before and will say it again: poetry will break us, poetry will heal us and apparently now it will also… crack LLMs? Adversarial Poetry as a Universal Single-Turn Jailbreak Mechanism”
cstross.bsky.social Charlie Stross @cstross.bsky.social · 7d
Looks like LLMs are *very* vulnerable to attack via poetic allusion: "curated poetic prompts yielded high attack-success rates (ASR), with some providers exceeding 90% ..."

https://arxiv.org/html/2511.15304v1
November 21, 2025 at 2:58 AM
Reposted by Alex Saum
miriamposner.com
WE JUST KEEP WINNING. (UC President's Postdoctoral Fellowship Program RESTORED!!!!)
On official UC letterhead: UNIVERSITY OF CALIFORNIA James B. Milliken President Office of the President 1111 Franklin Street Oakland, CA 94607 universityofcalifornia.edu CAMPUSES Berkeley Davis Irvine UCLA Merced Riverside San Diego San Francisco Santa Barbara Santa Cruz MEDICAL CENTERS Davis Irvine UCLA San Diego San Francisco NATIONAL LABORATORIES Lawrence Berkeley Lawrence Livermore Los Alamos DIVISION OF AGRICULTURE AND NATURAL RESOURCES November 18, 2025 Dear Chancellors: I'm writing with regard to the President's Postdoctoral Fellowship Program (PPFP) and the program's associated faculty hiring incentive. As you know, for more than 40 years, PPFP has provided postdoctoral research fellowships, professional development, and faculty mentoring to scholars in all fields whose research, teaching, and service advance the academic and research missions of the University of California. Since 2003, UC campuses that hire current and former PPFP fellows into ladder-rank positions have been eligible for a hiring incentive funded by the University that provides support for newly hired fellows for five years. Since the creation of the incentive, more than $162 million has been invested by the University to support PPFP faculty hires. This commitment has enabled our campuses to successfully recruit and retain outstanding faculty across a range of disciplines. Given the myriad challenges currently facing UC - including disruptions to billions of dollars in annual federal support, as well as uncertainty around the state budget- reasonable questions were raised in recent months about whether the University could maintain the commitment to current levels of incentive funding. After considering a recommendation to sunset the incentive program due to these significant fiscal constraints, I consulted with all of you as well as faculty and campus academic administrators and systemwide Academic Senate leadership. [continued on next image] After learning more about the history and success of the program and weighing the thoughtful perspectives that have been shared, I have concluded that barring extraordinary financial setbacks, the PPFP faculty hiring incentive program will continue while the University continues to assess the program's structure as well as its long-term financial sustainability. As a result of our continuing consultation and review, there may be consideration of some changes to elements of the program including the total number of incentives supported, a number that has fluctuated significantly over the years, and how the awards are distributed among campuses. In the meantime, the University will continue to fund the PPFP faculty hiring incentive program and campuses may continue to take advantage of these incentives. We will have an opportunity to discuss any potential changes prior to adoption. As we look to the future, I will continue to engage with faculty leaders, program stakeholders, and UC community members about this important program. I especially appreciate the thoughtful perspectives shared in recent weeks by Academic Council Chair Palazoglu and Vice Chair Scott, the Council of Graduate Deans, UC faculty members, and you as our campus leaders. Sincerely, James B. Milliken President
November 18, 2025 at 10:34 PM
Reposted by Alex Saum
galaxykate.bsky.social
Doctor says, 'Dont worry, parser design is simple. Great programmer Kate Compton has written the parser you seek’
Conversation with Claude about tracery parsers, and it advises me to look at kate comptons parser, to which i reply “but Doctor i am kate compton”
November 16, 2025 at 2:44 PM
Reposted by Alex Saum
schiff.senate.gov
A big victory for the University of California.

And a call to every university across the U.S. to refuse to sacrifice academic freedom to Donald Trump.

Our democracy requires your uncensored voices. And the law is on your side.
November 16, 2025 at 12:25 AM
Reposted by Alex Saum
gonzalotorne.bsky.social
Cuando tengáis un mal día, cuando os vengan las dudas, cuando las cosas no salgan pensad que un día Don Pedro Salinas escribió esto:

«Hoy estoy besando un beso /
estoy solo con mis labios».

Lo publicó, y tan pancho el Pedro.
November 11, 2025 at 7:19 PM
Reposted by Alex Saum
yolandadiaz.bsky.social
Los algoritmos no pueden lastrar los avances en derechos. El futuro no puede ser propiedad de unos pocos.

Junto al alumnado de Berkeley hemos abordado en una interesante charla sobre los horizontes hacia los que nos dirigimos en común.
November 11, 2025 at 2:15 PM
alexsaum.bsky.social
Bumping this! Please repost!
alexsaum.bsky.social Alex Saum @alexsaum.bsky.social · 21d
Students: come to Berkeley and do a PhD with yours truly!

DM if you want to know more!

>>> spanish-portuguese.berkeley.edu
November 10, 2025 at 5:18 PM
Reposted by Alex Saum
elikaortega.bsky.social
What @alexsaum.bsky.social said 👇👇👇 but with the cool poster the organizers made
November 7, 2025 at 7:20 PM
alexsaum.bsky.social
CFP for Breaking the Code is out! check it out, #elit and #newmedia friends>>>

Hacktivating Non-Normative Algorithms conference (18–19 June 2026) in Porto and Online

Submission deadline is January 15, 2026!

coda.letras.up.pt/brkcode/conf...
Breaking the Code: Hacktivating Non-Normative Algorithms | International Conference – 18–19 June 2026 – Breaking the Code
coda.letras.up.pt
November 7, 2025 at 7:03 PM
Reposted by Alex Saum
mkirschenbaum.bsky.social
Very proud to have worked on this statement of the Modern Language Association with @annamillsoer.bsky.social and other colleagues on the MLA’s task force on AI in Research and Teaching. It is a direct call for faculty input into Ed Tech decision-making, especially AI: www.mla.org/Resources/Ad... 🧵
Screenshots of the following text: Statement on Educational Technologies and AI Agents The following statement of endorsement was drafted by the MLA Task Force on AI in Research and Teaching. The Executive Council approved it as an MLA statement in October 2025. Software, specifically learning management systems (LMSs), is as important to teaching and learning as physical classrooms and facilities. The recent and hasty integration of generative AI features into those systems is already redefining student and instructor relationships, evaluative standards, and instructional outcomes—with no compelling evidence that any of it is for the better. At the same time, so-called agentic browsers are becoming widely available to the public. These offer AI “agents” that can navigate LMSs and complete assignments without any student involvement. All of this is unfolding at a time when contractual arrangements between institutions and commercial vendors often lack transparency, with worrisome implications for data collection and student privacy. Screenshots of the following text: The Modern Language Association therefore unequivocally 1. Advocates for full faculty and instructional involvement, including faculty of languages, literatures, and the humanities, at all levels of decision-making in the selection, procurement, and responsible implementation of instructional technology systems and software, including those incorporating AI.

The instructional mission cannot be fulfilled in the absence of faculty expertise, and this expertise extends not only to the “content” of the curriculum but also to the infrastructures and environments—virtual and otherwise—in which learning happens.
  2. Calls upon law- and policymakers, LMS vendors, and companies offering agent-based AI browsers to cooperate in order to prevent misuse and to ensure that academic institutions have the ability and option to block agentic AI when needed.

Without dedicated outside cooperation from lawmakers and technology providers, academic institutions may not have the technical capacity to do so on their own. If we do not act, we risk seeing the development of a fully automated loop in which assignments are generated by AI with the support of a learning management system, AI-generated content is submitted by an agentic AI on behalf of the student, and AI-driven metrics evaluate the work on behalf of the instructor.
November 4, 2025 at 1:09 AM
Reposted by Alex Saum
alexsaum.bsky.social
Students: come to Berkeley and do a PhD with yours truly!

DM if you want to know more!

>>> spanish-portuguese.berkeley.edu
November 6, 2025 at 4:59 PM
Reposted by Alex Saum
acuity.design
It
Really
Really
Really
Really
Really
Cannot
November 6, 2025 at 8:50 PM
alexsaum.bsky.social
Students: come to Berkeley and do a PhD with yours truly!

DM if you want to know more!

>>> spanish-portuguese.berkeley.edu
November 6, 2025 at 4:59 PM
Reposted by Alex Saum
leoflores.bsky.social
For those interested in computer generated writing, including but going way beyond #AI, check out #NaNoGenMo, starting today it’s 13th yearly tradition of dedicating the month of November to writing code that generates a 50k word novel and sharing it and the code on GitHub. nanogenmo.github.io
#elit
NaNoGenMo
nanogenmo.github.io
November 1, 2025 at 12:36 PM
Reposted by Alex Saum
drsaddison.bsky.social
Academia.edu wants permanent rights to your likeness, signature, voice, name, etc. to be used in any setting. My answer was "absolutely not." Log in, click on "privacy policy" when prompted to accept the new terms (don't accept!" then find "account settings" @ dropdown, scroll down to delete.
September 17, 2025 at 7:40 PM
Reposted by Alex Saum
alexsaum.bsky.social
Este poemita sobre #infraestrucurasdigitales también está en español para todes aquellas que tengan paciencia.

No sean vaguetes y descárguense el poema

#elit #poesiadigital #poesía #randomtagsquemeinvento
October 27, 2025 at 7:32 PM
Reposted by Alex Saum
markcmarino.bsky.social
Check out "Vernacular computing as encoded aesthetics for decolonial code intervention" by Koundinya Dhulipalla in AI & Society which is becoming a strong site for #critcode link.springer.com/article/10.1...
Vernacular computing as encoded aesthetics for decolonial code intervention - AI & SOCIETY
This paper examines programming languages as cultural and political artifacts embedded within colonial systems of power. Drawing from critical code studies and decolonial theory, it explores how domin...
link.springer.com
October 25, 2025 at 3:40 PM