Staff writer @TheAtlantic. Host of "Plain English" podcast @Ringer. Mondays on NPR's @hereandnow. Co-writing a book about progress.
18 added to My Authors
Mar 24 • 7 tweets • 4 min read
I'm really interested in the ability of AI like Midjourney to amplify artistic intelligence.
I've always been terrible at visual arts. Can't draw, can't paint, no instinct for design, or colors. Total black box for me.
So, I thought: Could Midjourney help me w/ interior design?
I really like leather-and-whiskey man cave vibes.
But I have no idea how to illustrate that, or describe what I like in detail.
So I gave Midjourney really basic instructions and it made this.
Mar 22 • 4 tweets • 2 min read
New study: When superhuman AI beat human Go players, professional Go players adapted by making "significantly better decisions"
Conclusion: "... Development of superhuman AI programs may have prompted human players to break away from traditional strategies"
This is all larval and nobody has any idea what's going to happen, but. One possibility is AI will serve for some as a cheat sheet (replacing intelligence) and for others a cognitive co-pilot (complementing intelligence), as @cwarzel and I discuss here open.spotify.com/episode/2KEFXK…
Mar 21 • 5 tweets • 2 min read
To kick off today's pod with @cwarzel on his great GPT-4 reporting, I spent some time in the open discussing three implications of generative AI:
1. A ceiling-raiser for collective intelligence 2. A Star Trek replicator for content 3. A human manipulator
open.spotify.com/episode/2KEFXK…
When I first wrote the intro, I thought I was writing about 2 good things (it's acing tests! it's a creativity spigot!) and 1 bad thing (it can bribe TaskRabbits!).
Then I talked to @LauraMartin531 about it and decided the good things could be bad and the bad thing could be good
Mar 18 • 4 tweets • 2 min read
This is the chart that has launched a thousand takes.
The conclusion seems obvious: Everything the govt touches goes to infinity in price. Everything pure capitalism touches goes to zero.
But there's another way of looking at this graph that dramatically changes the story.
As @mtkonczal finds, if you expand your analysis to 62 CPI categories and track their price changes in the 21st century, you get a slightly different story.
New paper: "Negativity drives online news consumption"
Blended study of 105,000 headlines and 370 million impressions concludes "each additional negative word [in a headline] increased the click-through rate by 2.3%"
"We find that news headlines containing positive language are significantly less likely to be clicked on. For a headline of average length, the presence of positive words in a news headline significantly decreases the likelihood of a headline being clicked on, by around 1.0%."
Mar 10 • 11 tweets • 4 min read
Silicon Valley Bank thread.
This is a dire overview of the implications. Many tech CEOs rely on SVB for payroll—and often for their own individual wealth management.
Consider 3 barriers to having as many kids as one would like:
- cost: infants are freaking expensive
- time: ... also, so fussy!
- distance: 2-income hhlds often sacrifice one partner's career bc they can't work in 2 cities at once
Mar 4 • 5 tweets • 2 min read
We'll see how this goes, but I'm concerned we're replacing one metric that's skewed toward rich kids with a bundle of metrics—GPA, extracurricular opportunities, fancy internships, clarinet mastery, soccer travel teams—that are even more stratified by SES status
But the justification doesn't make sense on its face.
"Individuals cannot be defined by any single factor"-> Exactly, so why reduce the number of factors by which to judge them?
Mar 3 • 5 tweets • 2 min read
I wrote about the lab leak and mask-mandate debates, what I make of The Truth in both cases, and how the demands of antagonistic news and social media consistently bungle genuinely confusing and deeply uncertain questions of science.
But its studies mostly ask: What happens when you randomly give ppl masks and masking info?
The answer: They mostly don't wear masks properly and fib to researchers.
Mar 1 • 7 tweets • 3 min read
This is a really interesting essay by @mattyglesias on the possibility—highlighted by several studies—that liberal teens are considerably more depressed than moderate or conservatives.
slowboring.com/p/why-are-youn…
As a liberal, I think it’s important to leave oneself open and vulnerable to the existence of injustices in the world.
But a political movement that, possibly thru its promotion of pessimistic rhetoric, leaves its adherents *unusually despairing & hopeless* has room to improve.
Feb 16 • 6 tweets • 3 min read
What is happening to youth mental health?
- 6 in 10 teen girls say they're persistently sad or hopeless
- share of girls who say they've contemplated suicide increased 50% in the last decade
- 25% of LGBQ teens say they've attempted suicide
We're witnessing a steady, decade-long increase in youth distress, which the pandemic clearly exacerbated.
Feb 14 • 5 tweets • 3 min read
The Youth Risk Behavior Survey is the gold standard for measuring teen attitudes.
The latest report was released this week.
The mental health crisis among teenagers—especially girls and LGBTQ teens—is getting worse, at an accelerating rate.
"Close to 70% of LGBQ+ students experienced persistent feelings of sadness or hopelessness during the past year and more than 50% had poor mental health during the past 30 days. Almost 25% attempted suicide during the past year."
Feb 1 • 5 tweets • 2 min read
This is the Super Bowl of all-time terrible Y axes.
Who ya got?
To me, there’s no question.
Truncated Y Axis has flair and style, but you’re just not going to find a more perfect(ly horrible) way to represent percentages than horizontal bars adding to 100% with a power log axis.
Feb 1 • 4 tweets • 1 min read
Hard to debate the central premise here: Trump-led Republicans have underperformed in three consecutive elections in part because they're still deluded by a 2016 black swan event into overestimating the public's taste for loud obnoxious jerks.
theatlantic.com/magazine/archi…
One of the Democrats' main challenges is that they don't really have a clear national base (despite over-performance among, e.g., Black Americans) and one of Republicans' main challenges is that they DO have an identifiable national base, and it's crazy.
Jan 31 • 4 tweets • 2 min read
This year, Pew asked US parents: What accomplishments or values are most important for your children as they become adults?
~9 in 10 parents named “jobs or careers [our children] enjoy” as a top value—more than "be honest" and 4x more than getting married or having kids
In 2019, Pew asked Americans: What is essential for a good and fulfilling life?
"Having a job or career they enjoy" beat everything else—not just having children, or being married, but also being in a committed relationship.
Jan 21 • 5 tweets • 3 min read
Everything is obvious once it's happened. But Frank, the student aid assistance startup acquired by JPM for $175m, seems like the most obvious con I've ever seen.
Fake email list ✅
Mathematically impossible customer claims ✅
Refusal to name its (fake) business partners! ✅
This is just basic addressable-market arithmetic!
It's like a boutique US food startup conned Goldman by claiming it sold hamburgers to 500 million Americans in the last five years, and the bankers forgot to Google the population of the United States.
Jan 20 • 4 tweets • 1 min read
The more I think about what this class of weight-loss drugs can do, and how it will change U.S. health care and self image, the more I think this might be the most under-covered story in the U.S. right now.
3 Qs asked and answered in today's pod—>
open.spotify.com/episode/2Xh2uR…
Morgan Stanley projects that if MDs prescribe obesity meds at same rate they medicalize hypertension, obesity drug market will be $90b/year by 2030.
That’s 30% of *total* US drug spending!
How's that gonna work?
What pressure will Congress/insurance feel to pick up the tab?
Jan 12 • 4 tweets • 2 min read
New report: US cancer mortality rate fell by ONE THIRD in the past three decades!
Key findings: Mortality declines are
- accelerating for lung cancer
- slowing for breast cancer
- stabilizing for prostate cancer
Key charts—>
I feel like a massively underrated story of public health progress is the decline of stomach cancer.
In 1930s, it led all cancers in death rate. In the last 100 years—with declines in smoking, improvements in food and water quality to eliminate bacteria—it's fallen almost 90%.
I wrote about the new blockbuster paper showing disruption is plunging across almost every domain in science, why it's happening, and why the rules of modern research might be creating a surfeit of crap
Some diff factors behind decline of "new ideas"
1) Burden of Knowledge: Harder to figure out new things in mature domains with many already-discovered facts
2) Burden of Volume: Modern Science™️ demands a high volume of papers in a way that might reduce quality of exploration
Dec 9, 2022 • 4 tweets • 2 min read
Not since peak Trump has anybody come close to Musk's power of discourse-gravity on Twitter. My entire feed is bent toward debating layperson vs congressional vs TOS definitions of "shadow-banning." It's pretty wild to observe this level of assignment-editor power.
Anyway, my take is Twitter's previous regime did a bad job, ran a money-losing firm & played at a hard game (fair and rigorous content moderation) poorly. The new regime is a cyclone of chaos that actively roots for better posting and electoral outcomes for the right. 🤷♂️
Dec 8, 2022 • 5 tweets • 3 min read
New big piece from me: The most important scientific and technological breakthroughs of 2022—from AI to deep space, new obesity drugs to mind-blowing biology experiments
Genuinely some of the most inspiring and interesting reporting I’ve done all year.
You all know about ChatGPT, Stable Diffusion, and the generative AI boom. Here are some more under-the-radar breakthroughs
(1) I'm not sure people realize that Yale scientists injected dead-for-an-hour pigs with a custommade substance that ....brought their organs back to life?!