James Hennessy

0 followers · 5 following · 273 views

on the atlas — 34

highlights — 73

For many families like my own, the Expressway symbolises a broader story of environmental injustice in the area, now dubbed ‘asthma alley’ due to its disproportionately high rates. Incidences of childhood asthma in the Bronx still rank 40 per cent higher than the New York City average, a fact attributed to elevated concentrations of particulate matter in the air. Last year, New York City’s mayor Eric Adams declared that the Cross-Bronx Expressway nurtured these inequalities by fragmenting largely Black and Latino working-class neighbourhoods while generating significant air pollution that has …
How the New York of Robert Moses shaped my father’s health | Aeon Essays
We lost my dad last year: the denouement in a courageously fought cancer battle that spanned more than two decades. I was 24 when he died—not as young as I could have been, but not old enough to negate a dull, almost-always-there sense of missing something. He was unpretentious, unfashionable, unfailingly reliable. He was corny and funny and sentimental. He was a rare combination of impossibly hard-working and deeply empathetic: a respiratory therapist for many years, he was an asthmatic who helped people breathe. We won’t ever be able to say for certain whether his lifelong lung issues, and l…
Sick City - Longreads
Second, that there are so many people in the world who want to understand physics but are unable to study it formally in a university setting for any number of reasons. These same people are very serious about learning physics, and not for any career purposes but simply because they want to understand the universe, and they are and have been dreadfully underserved and underestimated by the academic physics community (who do not take them seriously because they aren’t studying at colleges and universities) and by the authors of contemporary and popular physics books and the publishers of those …
Physics — Susan Rigetti
Many readers’ reaction to The Last Samurai focused on the bond between mother and son and the peculiarity of the son’s gifts. These things reflected only part of DeWitt’s purpose and her range. What was more rarely said was that The Last Samurai was also an angry book, about the curse of living with too much curiosity and intelligence in too pinched, prosaic, and conventional a world.
Your Name Here
“What problem do I actually have?” Because here’s the thing: When we can’t find a solution, it’s often because we’re trying to solve the wrong problem. In this newsletter, I’ll help you find the problem you actually have — by distilling it down to one simple word. Then we can find a solution that actually works. Let’s start with an example. A young professional recently came to me with a problem. He works at an advertising agency, and his ideas are always getting shot down. He’s frustrated, demoralized, and can’t figure out what’s wrong. “Should I just start executing my ideas without asking p…
How to Solve Your Big Problems, by Solving Your *Real* Problem
With the inaugural QWERTYFEST, Krulos and co-organizer Molly Snyder have worked to correct that error. Held from June 23 to June 25, QWERTYFEST was an inky paradise for collectors, artists, history buffs, and word nerds alike. The event featured live music, typewriting workshops, typewriter poetry ‘busking,’ and a local market featuring honey made by cemetery bees and a typewriter autographed by well-known typewriter enthusiast Tom Hanks. There was also brunch, open mics, and a scripted “Clackathon” performance. Developed alongside local theater company Quasimodo Physical Theater, the Clackath…
How Milwaukee Is Celebrating the Typewriter’s Long, Local History - Atlas Obscura
In this essay, I begin by identifying the reasons that automation has not wiped out a majority of jobs over the decades and centuries. Automation does indeed substitute for labor—as it is typically intended to do. However, automation also complements labor, raises output in ways that leads to higher demand for labor, and interacts with adjustments in labor supply. Journalists and even expert commentators tend to overstate the extent of machine substitution for human labor and ignore the strong complementarities between automation and labor that increase productivity, raise earnings, and augmen…
Why Are There Still So Many Jobs? The History and Future of Workplace Automation - American Economic Association
Unfortunately, it’s not immediately clear what replaces the atoms of the periodic table in the standard model. Are the fundamental building blocks of reality quantum particles, quantum fields, or some combination of the two? Before tackling this difficult question, let us consider the debate between particles and fields in the context of a classical (non-quantum) theory: Maxwell’s theory of electrodynamics.
Is everything made of particles, fields or both combined? | Aeon Essays
When Elias Stadiatis descended into the indigo-blue water, he had a normal day of searching for sponges ahead of him. Weighed down in a copper diving suit, surrounded by a tangle of breathing tubes, Stadiatis eventually reached the seafloor. As he squinted into the dim, he took in a haunting scene: all around were the fuzzy outlines of human body parts. As he surfaced in a pool of bubbles, he frantically informed the captain he'd found a heap of rotting corpses.
How many shipwrecks are there in the world's oceans? - BBC Future
Of course not all problems can be solved so easily, but that doesn’t make it a poor use of the leadership time to engage. Maybe the problem was harder than we realized, maybe it required a different skill set, maybe it needed more authority; whatever the problem, it is good to be open with it so everyone can set expectations appropriately. In an ideal world everyone would be operating under conditions that allowed them to succeed on their own, but until we manage to create such a world loop you should do whatever you must to get it done, even if that means asking for help.
Get It Done
Ship Vision First, Final Product Later TL;DR: Ship Vision First, Final Product Later
Why You Can't Kickstart a Conference - Abner
There’s not much to add here: people are more willing to support you if they know this is how you’ll feed your family. If you have a well-paying job elsewhere, and this conference is a profit-making vehicle and not your livelihood, well… there’s nothing wrong with that but it’s only natural it will weaken your audience’s resolve.
Why You Can't Kickstart a Conference - Abner
While I appreciate the transparency, I find myself perplexed. Why not have the speakers sign a contract? That’s how you state with confidence that your speakers will, without a doubt, attend. Yes life gets in the way, but we can all agree that a legally-binding agreement (ideally with compensation / clawback provisions) suddenly eliminates these “life problems” in 99% of cases 🙂 The remaining 1% becomes an inconsequential risk not worth mentioning.
Why You Can't Kickstart a Conference - Abner
Sponsorship deals of any kind are the reason the software industry is in the sorry shape it’s in. To the programmer not privy to backroom shenanigans this pronouncement may seem a little out there, but it’s the reality: tech conferences have the power to shape what we as software engineers focus on (a single talk at GDC can alter the trajectory of the game dev industry.) As of right now, nearly all events are sponsored by the tech giants.
Why You Can't Kickstart a Conference - Abner
Theme #1: Unbounded Social Challenges - Who would willingly entrust you with the operation of a fancy venue in a major American city? Should you manage to secure such trust, congratulations—you now bear responsibility for the caterers, ushers, security guards, A/V technicians, janitors, booking agents, and the insurance company (because you better get insured.) The venue owners will hold you liable for any incidents transpiring within premises. I hope you haven’t forgotten about your speakers and your ticket holders, who rely on your organization. How about marketing to strangers so you can ac…
Why You Can't Kickstart a Conference - Abner
As a full-time indie organizer—who is currently doing okay for himself—I hesitated to write this essay because it could seem like I’m punching down. That said, I believe there are significant lessons to mine from this Kickstarter initiative: enough of them to be worth coming off as ruthless.
Why You Can't Kickstart a Conference - Abner
Now researchers have discovered how skates evolved their distinctive profile: Rearrangements in the skate’s DNA sequence altered the 3D structure of its genome and disrupted ancient connections between key developmental genes and the regulatory sequences that governed them. Those changes in turn redrafted the animal’s body plan. The scientists reported their findings in Nature in April.
How 3D Changes in the Genome Turned Sharks Into Skates | Quanta Magazine
We think this should call into greater question the recent flurry of academic work using Large Language Models (LLMs) like GPT to shortcut data validation — a foundational principle in any kind of science, and especially machine learning. These papers are often uploaded to Arxiv and widely shared on Twitter before any legitimate peer review. In this case, potentially spreading bad information and setting a poor precedent for future work.
No, GPT4 can’t ace MIT
A paper seemingly demonstrating that GPT-4 could ace the MIT EECS + Math curriculum recently went viral on twitter, getting over 500 retweets in a single day. Like most, we were excited to read the analysis behind such a feat, but what we found left us surprised and disappointed. Even though the authors of the paper said they manually reviewed the published dataset for quality, we found clear signs that a significant portion of the evaluation dataset was contaminated in such a way that let the model cheat like a student who was fed the answers to a test right before taking it.
No, GPT4 can’t ace MIT
AI in Bio take three forms: in silico drug discovery, computer vision and phenotypic drug discovery, and generative AI predictions
Gmail
CCS has always been a controversial technology, with many critics seeing it as merely a tool by the fossil fuel industry to delay real climate action. Yet, in the debate, a key argument is that CCS will be required anyway for some industrial emissions – and cement is the prime example. CCS for Cement should be less controversial than for Fossil Fuels It would therefore appear that a cement plant with CCS would be much less controversial than other CCS projects. However, despite cement's key role in the CCS debate, not a single cement plant with CCS is operational today, and only one is under c…
Cement's future could be a combination of Carbon Capture and Electrification - Industry Decarbonization Newsletter
Our AI red team is a cross-functional team made up of offensive security professionals and data scientists. We use our combined skills to assess our ML systems to identify and help mitigate any risks from the perspective of information security.
NVIDIA AI Red Team: An Introduction | NVIDIA Technical Blog
Joann Berman: The thing about San Francisco that made it different from New York and LA was that it came from a visual art background because of the Art Institute. Ruby Ray: The first Search & Destroy address, 3426 Jones, was an apartment that was right across the street from the Art Institute. There was a total interaction between the Art Institute and the punk scene. Sally Webster: A lot of people were at the Art Institute. Bruce Pollack, who had another sort of performance space around the corner called the A-Hole. And Sue White. Jill Hoffman from Target Video. I went to the College of Arts…
The Quietus | Features | Tome On The Range | Punk, Art & Anti-Art: How The Late 70s SF Scene Hit Different
In principle, we should not touch our test set until after we have chosen all our hyperparameters. Were we to use the test data in the model selection process, there is a risk that we might overfit the test data. Then we would be in serious trouble. If we overfit our training data, there is always the evaluation on test data to keep us honest. But if we overfit the test data, how would we ever know? See Ong et al. (2005) for an example how this can lead to absurd results even for models where the complexity can be tightly controlled.
3.6. Generalization — Dive into Deep Learning 1.0.0-beta0 documentation
Fano had solved this part of the problem. He realized that he could use codes of varying lengths without needing costly spaces, as long as he never used the same pattern of digits as both a complete code and the start of another code. For instance, if the letter S was so common in a particular message that Fano assigned it the extremely short code 01, then no other letter in that message would be encoded with anything that started 01; codes like 010, 011 or 0101 would all be forbidden
How Lossless Data Compression Works | Quanta Magazine
The Rao-Blackwell theorem states that if we have an estimator for a metric and a sufficient statistic for the data, we can improve the estimator by conditioning on the sufficient statistic. We examine this improvement is in terms of mean squared error, which is a measure of the estimator’s accuracy.
Understanding the Rao-Blackwell Theorem
Also by contrast: things like universal assemblers remained relatively somewhat vague as a goal; MNT didn't have a particular killer app that a specific really big customer desperately wanted; and as a result top labs weren't explicitly working on MNT as a goal.
Sporadica
A question I've wondered a fair bit about: why did quantum computing take off somewhat smoothly, while molecular nanotechnology did not? There's many similarities: Both were proposed in the early-to-mid-1980s. Both had precursor work done by Feynman, although the key papers were written by others. Both were exciting, but a little vague. The quantum computing model got firmed up pretty quickly into a very concrete goal; things like universal assemblers remained (and remain AFAIK) somewhat vaguely specified.
Sporadica
Treating ChatGPT as a Google-quality search engine for facts is a small-scale version of “enchanted determinism,” and one much less damaging than the kind practiced by some politicians and bureaucrats (not to mention A.I. capitalists). As a specific phenomena, the lawyer citing fake cases or the student asking about imaginary articles seem less like heralds of some wacky new world and more like victims of the difficult process through which new technologies are adapted by people and institutions. The hallucinations are not in and of themselves “the problem”--how people receive, process, and un…
I cannot believe the shit that morons are getting up to with ChatGPT
This discomfort was brought to the interim leadership group as a problem that must be fixed immediately. There was no discussion of the ramifications of making a change at this point. There were complaints that this pushback was not sufficient to require change, but this feedback was ignored. The discussion focused on the discomfort of the team members and changing JeanHeyd away from being a keynote speaker as the only solution.
Why I left Rust
Since I’m on the topic of Max Weber, religion and technology already, here’s a half-developed theory of Elon Musk that I’ve been nurturing for a while. I’ve trotted it out informally at a couple of meetings, and I’m not completely convinced it is right, but it’s prima facie plausible, and I’ve gotten some entertainment from it. My argument is that Musk is doing such a terrible job as Twitter CEO because he is a cult leader trying to manage a church hierarchy. Relatedly – one of SV’s culture problems right now is that it has a lot of cult leaders who hate the dull routinization of everyday life…
The Cult of the Founders — Crooked Timber
positions at Carnegie Mellon and Stanford before that. Brooks, who is now working on his third robotics startup, Robust.AI, has written hundreds of articles and half a dozen books and was featured in the motion picture Fast, Cheap & Out of Control. He is a rare technical leader who has had a stellar career in business and in academia and has still found time to engage with the popular culture
Just Calm Down About GPT-4 Already - IEEE Spectrum
Vaclav Smil is a theory-builder, but rarely for historians, he’s pretty good at it. One promising sign is that he has a cult following (Bill Gates once said he waited for new Vaclav Smil books the way he waited for Star Wars movies), but I actually think the reason his approach works is that he’s chosen the most impersonal and non-negotiable of all qualities to build his theory around: energy constraints. For Smil, the history of human civilization is nothing more and nothing less than the history of attempts to balance a single equation: joules of energy input, in the form of food and fuel, v…
REVIEW: Energy and Civilization, by Vaclav Smil
Matty understands on an intrinsic level what a lot of ‘internet children’ feel: this indescribable feeling of hyper-self-aware cringe that creates a hard shell around the deep-rooted desire to be earnest and genuine
Taylor Swift and Matty Healy are reportedly dating, and fans are furious - Vox
eye-to-eye communication, nothing is made personal and the debate involves a solution
The Tao of Spo: Erik Spoelstra’s compassion, competitiveness and confrontations - The Athletic
“The pain of losing put a steeliness in him. His overall presence in the gym became bigger and stronger. His willingness to go at people and really challenge them grew,” Fizdale said. “It was a dramatic shift … because when he came back the next year, man … he really had an oomph to him. He really held our team to a high standard. He did not waver.”
The Tao of Spo: Erik Spoelstra’s compassion, competitiveness and confrontations - The Athletic
The playbook wasn’t the only thing that changed. So too did Spoelstra. He had tried so hard in his first four seasons to sound like a head coach, carry himself like a head coach, that he never got around to being a head coach. But at some point during the longest night of his life, he found himself and found his voice.
The Tao of Spo: Erik Spoelstra’s compassion, competitiveness and confrontations - The Athletic
In the fall of 2012, people in the surrounding tents began to keep an eye on Koyama-san, who spent most of her time alone. Someone came to tell me that she had collapsed and that they were worried about her. When I went to check on her, I heard a voice from inside the tent saying, “I’ll be fine. I’m just a little tired.” I handed her some food through a gap in the tent, and in exchange she gave me what looked to be a small, handmade item tied up with sparkly buttons. I didn’t know what it was, but for a while after that, Koyama-san would occasionally give me one of these sparkly objects.
A Path That Leads to Heaven | Koyama-san Notebook Workshop
Carefully wrapped in layers of plastic to protect against the rain and humidity, the dozens of small notebooks that Koyama-san left behind seem as though they’ve been waiting to be read by someone. As I flip through them, I realize my intuition is probably correct. I don’t know whether to call them autofiction or a diary, but these little notebooks have captivated all of us, despite our different backgrounds.
A Path That Leads to Heaven | Koyama-san Notebook Workshop
This paper is extremely similar in domain (same corpus, etc - but that's not surprising since everyone uses these) but they're leaning heavily on the pretraining allowing capabilities towards few and zero shot, which is already well understood. Ultimately I think it's a good resource to use for the code, and if the API ends up being easier to change some of the internals on compared to the many options out there such as scispacy, or any of the pipelines used to achieve Pubtator, then it's a welcome addition.
A transformer-based method for zero and few-shot biomedical NER | Hacker News
The forum was structured in an analogous way to how Democratic candidates’ primary-season town halls worked in 2020: which is to say, those were in front of an audience of Democrats and Democratic-leaning voters, eager to hear the sorts of base-pleasing messages that define a primary campaign. We can argue about the best ways to cover political candidates, but I see value in these formats. The people in the audiences are representative of the people who will vote in primary elections, so their attitudes are relevant, and when they ask questions they actually tend to be more focused on policy (…
Gmail
Note that the divergence operation is independent for each I, and this is great because it means we can parallelize it! This is where KA comes into place. To be able to generate the divergence operator using KA, we need to write the divergence kernel which is just the divergence operator written in KA style.
Porting WaterLily.jl into a backend-agnostic solver - Bernat Font
Neither Gabriel García Márquez nor Mario Vargas Llosa had yet been born when the Guatemalan Miguel Ángel Asturias began to write his first novel, El Señor Presidente, in December 1922. He labored on it for a decade while living in self-imposed exile in Paris, then returned home when the Great Depression left him strapped for money, only to find that his work was unpublishable because the dictator whose reign it portrayed had given way to an even more cruel and oppressive one. When he finally self-published the novel in Mexico in 1946, it was riddled with typographical errors, and a definitive …
The Inventor of Magical Realism | Larry Rohter | The New York Review of Books
σ = ∰ ( ∇ ⋅ → u ) d V = ∯ → u ⋅ ^ n d S → σ i , j = ( u i + 1 , j − u i , j ) + ( v i , j + 1 − v i , j ) ,
Porting WaterLily.jl into a backend-agnostic solver - Bernat Font
v1.0 has ported the solver from a serial CPU execution to a backend-agnostic execution including multi-threaded CPU and GPU from different vendors (NVIDIA and AMD) thanks to KernelAbstractions.jl (KA). In this post, we will review our approach to port the code together with the challenges we have faced.
Porting WaterLily.jl into a backend-agnostic solver - Bernat Font
Data parallelism: Split the data among (possibly model-parallel) replicas of the model
Transformer Math 101 | EleutherAI Blog
Despite Brain’s tremendous value creation from its early funding of open-ended ML research, it is becoming increasingly apparent to Google that it does not know how to capture that value. Google is of course not obligated to fund open-ended research, but it will nevertheless be a sad day for researchers and for the world if Google turns down its investments.
Why does did Google Brain exist?
That’s how I’ve felt watching the news about solar geoengineering unfold over the past several months. Thanks to a few big news events, efforts to cool down the planet by reflecting some sunlight back into space are suddenly a huge part of the public conversation.
The inside scoop on solar geoengineering | MIT Technology Review
Ms Van Wingerden’s Floating Farm is the apotheosis of centuries of Dutch thinking about how to grow lots of food in a crowded corner of northern Europe. Since the age of Rembrandt and Vermeer, land has been reclaimed from the sea and windmills erected to drain the plains. Town-size greenhouses are built to grow tulips or vegetables. A food shortage during the second world war convinced the Dutch they needed to grow as much as their fields could manage. Calvinist industriousness turned the Netherlands into an unlikely agrarian powerhouse: with more than €100bn ($108bn) of annual farming sales o…
The cucumber Saudis: how the Dutch got too good at farming | The Economist
We will not investigate the sources of this overhead in this blog post and leave it to other posts or locations for now, instead focusing on memory for model training in the rest of this post. If you’re interested in learning more about the calculations required for inference, check out this fantastic blog post covering inference in depth. Now, on to training!
Transformer Math 101 | EleutherAI Blog