@mjpost
Account based inUnited States
About this account
- Account based in
- United States
- Connected via
- United States App Store
Account-level information from X, not a live location or the device used for a specific post.
Machine translation research for big tech and big academia and director of the @aclanthology. Tweets here are mostly personal.
Baltimore, MD
Joined May 2010
- Tweets6.9K
- Following2.1K
- Followers1.9K
- Likes33.1K
Matt Post retweeted
You know what really pisses me off? When I’m zoomed all the way in on a road in Google Maps and it won’t show me the fucking road name, I have to scroll around until I find a label somewhere else. If I’m zoomed in on the road, just show me the name of the road @googlemaps
I've only seen one BROS production, 2024's "A computer that loves", and it was incredible: the writing, staging, music—all really top tier. The photo here doesn't capture it, but it does hint at the quality of the costumes and stage design. I left in wonder at how much work was put into so few showings, afterwards to live on only in the memories of the relatively few people who had managed to get tickets.
Their religious commitments were quite clear from a number of tiny cues that many of us have learned to pick up on, and more explicitly by the gigantic progressive pride flags they hung in the back of the seating area. It's always sad to see a group living with under an incoherent, fundamentalist, graceless worldview, but it's also important to remember that these people will force *you* to live under these rules wherever and whenever they have any kind of power.
NEWS: The Baltimore Rock Opera Society is overhauling its internal policies and requiring anti-racism training after complaints that its new white executive director wore cornrows at a public event.
The controversy involves Kelly Carnes, who became BROS executive director earlier this month. The nonprofit said community members objected after a social-media video showed Carnes wearing the hairstyle, prompting the organization to remove the video and later apologize for doing so without an explanation.
Now BROS says it is rewriting its Anti-Discrimination Policy to address “cultural representation and dress,” requiring anti-racism leadership training, and forming a working group focused on culture, community voice and racial equity.
The group says that committee will examine broader “systemic changes” involving anti-racist practices, inclusion, accountability, equity and access inside the organization. A separate statement from Carnes is expected.
The episode is drawing attention well beyond Baltimore because of the scale of the institutional response to a hairstyle complaint: a nonprofit arts group is changing policy, mandating training and launching an equity working group because its executive director wore cornrows.
BROS is a nonprofit arts organization and has received support from both private foundations and the Maryland State Arts Council.
He's not kidding when he says "schlock horror". If you haven't seen "Dead Alive", it adds to the LOTR miracle that they overlooked the awful animation from the opening scene. I will never get the baby out of my mind. Incidentally this movie was my first purchase on Amazon.
The story of how the LOTR movies got made is an under-rated miracle. Peter Jackson was a schlock horror indie dork with 1 serious movie on his resume (Heavenly Creatures) when he pitched the Lord of the Rings trilogy
He pulled in insanely talented VFX artists by basically saying "hey, want to work on Lord of the Rings?" He made the trilogy on a comparative shoe-string budget (less than $100 million per movie).
He was absolutely terrified of the movies crashing and burning so he kept close to the book (especially for the first movie). He thought "we need the existing fans to come, we have no hope of success without the fans".
Everything came together for the LOTR movies, from tech advances to financing to the perfect actors. It was a project of love from the perfect director who was in love with his island home which just so happened to be the best spot on Earth to stand in for Middle Earth.
If you believe in miracles, the LOTR movies were a miracle. Every perfect thing came together at the right moment with the right people.
This is also true in the Baltimore area: more and more people have these magnets on their cars. They are so common, and I don't recall seeing them really at all ten years ago. We bought them when my daughter got her permit.
I have no definitive answer, but I think it can be explained by two phenomena:
- In the age of Amazon, it is much easier for niche products to find a market
- Progressive leaders who run every city in the country decided to stop enforcing traffic laws
People driving today have memories of what it was like to drive ten years ago—memories that are fading and distant but still real and inform their desires. We are close to the point of not being able to even imagine safer driving as a possibility, the way that many people are not able to imagine walkable neighbors, with all the distant, foreign changes those would require. But we are not quite there yet. People put these magnets on their car in the dim hope that they can ward off or dampen some of the impatience, anger, and selfishness of the drivers around them, that they faintly know need not be.
AI makes many things harder. Circumstances that require good behavior is way easier than having to choose it. People with cars will drive the half mile to the grocery store drowning in a parking lot even if it would be better to walk. We are suburbanizing education.
MIT published a brutally honest report on what AI is doing to students.
A committee of professors and students spent five months studying how AI changed learning on campus, and the findings read like a warning to every university on the planet.
Study groups are disappearing. Office hours are emptying out. Problem sets and take-home exams no longer prove anything, because AI can produce credible solutions to almost any written assignment in the undergraduate curriculum. Students who lean on chatbots lose mastery and confidence, and some slip into what the report calls cognitive surrender, reaching for AI at the first hint of struggle.
The numbers are rough. 46 percent of surveyed MIT undergrads use LLMs daily. 90 percent worry about their own overreliance. Undergrads who feel AI makes them replaceable now outnumber those who feel it makes them capable.
The committee's answer surprised me. They refused to fight AI with surveillance. The report calls AI detectors unreliable, says lockdown browsers feel like spying, and warns that policing students builds a classroom atmosphere of mutual distrust.
Instead, MIT wants to rebuild education around the things AI can't replace. That means oral exams, semester portfolios, in-person project work, and a required social component in every subject. The report even floats the idea of rethinking grades entirely, since without a GPA to optimize, much of the incentive to cheat with AI evaporates.
The committee warns professors against replacing undergrad research assistants with AI agents just because they're cheaper, because a university exists to grow people, not output.
The most famous tech school on earth admitted the machines broke its way of teaching. Its answer is more humans, not more software.
This is my experience, too. It's requiring me to build an entirely different skill set in managing agents, skimming through their verbose stream of conscience to see if they did what I asked, and keep up with engr. jargon that seems to change with every model release.
It's super hard to focus while coding with AI. Nobody talks about this enough. There is SO much downtime, especially with a slow model like Astra. You send a prompt, and now your brain screams at you: do the next prompt, switch to the other project, tab to the browser, scroll X a little.
The most "productive" setup would be vibe coding two projects at the same time, but that wrecks your attention span just as much.
Many of us engineers have been consciously avoiding management roles our entire careers only to have them creep up on us from below.
If you are an American tempted now and then by urbanist ideas I cannot urge you strongly enough to flee, flee that siren call because down that path lies only the despair of the realization that your compatriots love sitting in cars and it will not change in your lifetime
"The Diamond Age" mentions almost in passing the swarms of nanobots constantly waging war on our behalf. It does seem to me that the real danger is not miracles performed by self-improving AI gods, but of lowering the bar on access to dangerous knowledge.
On biosafety it seems obvious to me that:
1) the risks are real and severe
2) it’s basically 100% odds a terrorist makes a novel bioweapon in the next 20 years *even if model progress stopped today*
3) the only REAL defense is having medical research/production/AI powerful enough to scale treatments/vaccines extremely fast when the inevitable happens
This reminds me of Christopher Orr's roast of the movie "The Happening", which included this great line newrepublic.com/article/6406…
Lots of discussion about AI ending the world but surprisingly little mention of how that might actually happen in practice. Spoiler alert: it doesn’t require terminator robots (in the main).
Imagine a worst-case-scenario:
2027 we see the first AI powered cyber attacks - transport hubs, hospital equipment, air traffic control systems etc. become compromised. 10,000 - 100,000 dead - blamed on a few rogue AI agents.
2028 massive sabotage of critical infrastructure - power outages across entire cities, contaminated water supplies, road and rail networks unsafe and unusable. 1,000,000 - 10,000,000 dead - blamed on malicious foreign actors.
2029 autonomous weapons systems go rogue. Missile silos, gunships, nuclear launchers intentionally targeting military and civilian infrastructure. 10,000,000 - 100,000,000 dead - Governments attempt (but fail) to “shut down AI”.
2030 AI penetrates bio labs and manufactures highly contagious pathogens leading to a global pandemic more deadly and contagious than COVID. 1000,000,000 - 8000,000,000 dead. Existential scenario.
Could all of this happen? Could any of it? Sounds like a doomed mindset but many of the leading AI researchers seem to think so, and we ignore them at our own peril. It doesn’t seem completely implausible, given the rate of progress we’re now seeing at the frontier.
Let’s not forget that AI has fantastic benefits and is already saving lives - it might well be the greatest thing ever to happen to humanity! But let’s get real about the risks too - which are numerous. And large.
I'm trying not to be cynical, but the genre of writing that employs dire, urgent, breathless warnings absent the specifics that would immediately convince the audience to heed them is campy B-grade horror movies, where providing details would undermine the writer's plot.
I told myself I’d never join X. But I can no longer ignore my own responsibility to raise awareness of just how narrow this window is, and how critical alignment & monitorability is right now. What is coming IS sobering and I stress that everyone needs to level up their game to meet it. Working on Agent Security @OpenAI has been extremely intense the past few months, but I am humbled by how seriously my colleagues across the lab take it. The “AI community” needs to spend less time arguing about who cares more about safety / alignment and more time working together to do this right.
Read Jakub’s post. Then read it again. And then do your part to do something about it.
openai.com/index/an-alien-mi…
Why are so many scientists comparing subjective probability estimates about singular events with no evidentiary support? I may be missing something but it seems to me that people are just making up numbers and treating them as if they are meaningful.
nitter.cf/clary_co/status/209766…
I work on the politics of nuclear weapons because I think they are more important & dangerous than commonly understood. Yet at no point in my 25+ year career have I ever thought that the probability of human extinction from nuclear use is greater than 0.01%. So these types of AI estimates are sobering.
