ทวีตที่ปักหมุด
Drew
97.1K posts
Drew
@onefiftyfivemm
bene vixit, qui bene latuit
Ix เข้าร่วม Temmuz 2009
1.2K กำลังติดตาม1.5K ผู้ติดตาม
@RussellShaw_MLS Further, you think of Claude as artificial intelligence, but the analogy to which you reach first is a centrifuge? A tool? This suggests very limited ethical and moral development.
English
@RussellShaw_MLS A rather prickly response to what turns out to be a correct observation, don't you think?
English
The lab has been doing exactly this for over a decade. We just didn’t call it AI.
Auto-verification rules in clinical laboratories release normal results to the patient chart without a human ever reviewing them. The system checks the result against predefined criteria: within reference range, no delta check flag, no critical value, no instrument error code. If all criteria pass, the result goes straight to the chart. No human eyes. No scientist review. Released automatically.
The criteria are narrow. The guardrails are strict. The system kicks abnormals, critical values, and flagged specimens to a human for manual review. Sound familiar? That’s exactly the framework Legion Health just described. Narrow set of stable cases. Predefined criteria. Automatic processing within the guardrails. Human escalation when the case falls outside the rules.
The difference is that auto-verification in the lab processes millions of results per day across every hospital system in the country and nobody blinks. An AI renewing a stable patient’s SSRI prescription using the same logic structure makes national news.
The question worth watching isn’t whether this works. It will. For stable refills of low-risk maintenance medications in monitored patients, the decision tree is simple enough that a well-designed system will perform reliably. The question is what happens when it works well enough that the guardrails start expanding. 15 drugs becomes 30. Stable patients becomes “mostly stable.” The pharmacist review becomes a spot check instead of universal review. That’s the trajectory I watched in lab automation. Every guardrail that was designed to keep humans in the loop eventually gets widened once the system proves it doesn’t need them. And it works fine until the one case where it doesn’t.
The narrow version of this is reasonable. The version it becomes in five years is the one I’d be watching.
Rohan Paul@rohanpaul_ai
This is just great. 👏👏 We just crossed a line in medicine. San Francisco startup Legion Health now is allowed to use an AI chatbot to renew certain psychiatric prescriptions without a doctor signing off on every case. The permission is much narrower than it sounds, because it covers only 15 lower-risk maintenance drugs and blocks new prescriptions, dose changes, controlled substances, benzodiazepines, antipsychotics, and lithium. The system is also fenced in around stable patients, and it must kick cases to humans for suicidality, mania, severe side effects, pregnancy, or any patient who asks for a person. So the experiment is not “AI writes whatever it wants with no humans involved,” and it is also not “doctors do everything and AI is just decoration.” It is a guardrailed handoff where the AI does the first-pass refill decision for a narrow set of stable psychiatric patients, and humans monitor it closely at first, then less often if it performs well. Legion Health’s system is not being asked to diagnose a crisis or invent a treatment plan from scratch. Reports say it can renew a narrow set of existing prescriptions, only for patients already stabilized by a human psychiatrist, with pharmacists and regulators still in the loop. Even so, psychiatry is unusually hard to automate because the decisive information is often not just what a patient says. --- nationaltoday .com/us/ca/san-francisco/news/2026/04/06/ai-psychiatry-startup-approved-to-prescribe-meds/
English
Drew รีทวีตแล้ว
The source of the data center water psychosis is @_KarenHao, whose book Empire of AI was a NYT best seller but overestimated water use by 100,000% (lol).
The response was just “oopsies” and all the incorrect books were kept in circulation 🤷🏻♂️
Jeremy Horpedahl 🥚📉@jmhorp
A typical data center uses about the same amount of water as a golf course, and the same amount of electricity as a steel plant. Yet for some people, they have become The Worst Thing In The World. Where does this motivation come from?
English
@onefiftyfivemm Well, thanks I guess. You sounded like you’re a fellow tech. Thought we might actually have a conversation.
English
Fair point and I’ll take it. The analogy is structural, not clinical. Both systems use rule-based decision trees with narrow criteria and human escalation for exceptions. That’s the parallel I was drawing. The logic architecture is similar even if the inputs are fundamentally different.
But you’re right that the inputs matter. In the lab, auto-verification is comparing a quantitative measurement against multiple layers of objective criteria. Not just reference ranges. Delta checks compare the current result against the patient’s own historical values and flag significant changes that might indicate a specimen mix-up, acute clinical change, or pre-analytical error, even when the result technically falls within the normal reference range. A hemoglobin that drops from 14 to 10 in three days is flagged not because 10 is critically low but because the delta from the patient’s baseline is clinically significant. That’s longitudinal, patient-specific, data-driven decision making built into the automation. Glucose is glucose. Hemoglobin is hemoglobin. The numbers are objective, the thresholds are validated, the delta logic is anchored to the patient’s own history, and the decision to release or hold the result is deterministic.
Psychiatry has no equivalent. There’s no serum serotonin level that confirms depression. No plasma norepinephrine threshold that validates an anxiety diagnosis. No biomarker that tells you the SSRI is working at the molecular level. The clinical assessment is subjective, context-dependent, and relies on patient self-report, behavioral observation, and clinician judgment. All of which are things a chatbot collecting text responses cannot replicate with the same fidelity.
So the structural analogy holds. Narrow criteria, guardrails, human escalation. But the epistemological foundation is different in a way that matters. The lab auto-verifies against biology. The AI would be auto-verifying against self-reported symptoms with no objective anchor. That’s a weaker foundation for the same confidence level. Conceded.
English
@onefiftyfivemm i don't believe this at all, based on my actual healthcare experience. they seem blithely unbothered if patients are upset with them
English
The modal MD isn't greedy, but the ones who work for and with the AMA are expressly and consciously greedy. That isn't your normal practicing physician.
The modal MD is a herd animal who thinks he's an apex predator who is plagued by bad dreams where one of his parents is upset
pnorm@paleonormie
if the AMA would allow our unis to actually train a sufficient number of doctors then doctor wouldn't be an exhausting 70 hour a week job. Everyone knows that you work hard, but nobody feels sorry for you because we all know it wouldn't be like that if you weren't so damn greedy
English
That's right: trains and gdp don't track together
You've discovered the phenomenon of independent sets, how exciting for you
Hayden@the_transit_guy
Iowa, an economy with a $257 billion GDP, sees fewer trains in a week than this mountaintop station 6,700 feet above sea level in Switzerland does in a day.
English
It appears we have our own energiewende death cult nucleus
Duncan S. Campbell@duncancampbell
In Pennsylvania this weekend: 1) Liberal family friends are very concerned about data centers. “What’s happening is horrible.” 2) Drove through very red rural areas for a few hours and “Stop Data Centers” signs were everywhere. This may be the most unifying issue in America.
English
日本からこんにちは🇯🇵👋アメリカ人に質問があります🇺🇸僕は本物の銃を見たことがありません。アニメや映画でしか見たことがありません。アメリカ人は本物の銃を見たことがあるんですか？それは普通のことなんですか？🤔
日本語
@mauddweeb Fork tongued dick has no self, so of course he is desperate to know whether other people approve of him listening to audiobooks
English
Why would I feel bad about an audiobook
Richard Hanania@RichardHanania
Illiad and Odyssey are the rare books where you don’t have to feel bad about listening to an audiobook version, because it’s closer to how people originally experienced them. But the reader of the version I have is a woman, so I worry that changes the experience.
English
Because it is not winning in California, because California is not ruled by policy or ideology but by the rivalries of ethnic gangs in the permanent bureaucracy who are, to a man, nihilists
Derek Thompson@DKThomp
Very important question: If YIMBY/Abundance is winning in California, why isn't the state building more homes? Brian Hanlon (@hanlonbt): "The main reasons we don’t have housing come down to three basic things." - First, zoning. It’s illegal to build dense housing even when there’s demand. - Second, streamlining and permits. Even if the housing is legal to build, getting approval can be time-consuming, uncertain, and expensive. - Third, costs: not just higher labor costs, but also govt-imposed costs, like inclusionary zoning and parks fees I would add: The most recent YIMBY wins have coincided with a high interest rate environment and a Trump immigration policy that especially affects places like CA that rely on foreign-born workers. That said, interest rates are national and CA's building crisis is special. asteriskmag.substack.com/p/if-yimby-is-…
English
@NemoOdysseus @EricRichards22 If she had been a determined anti communist rather than the other thing she would have already been lost to time, entirely
English
@EricRichards22 I figured that Keller endured through the 20c bc youth publishers wanted to fill the “kids need a disabled figure to read about” position. That’s how I first read about her.
English
There isn't really any reason why anybody should know about Helen Keller, or why it would matter whether or not she existed, or did the things attributed to her
unless you are making of her some kind of mytho-religious figure
It's only as valuable a thing to know as Dick Tracy
GIF
Reddit Lies@reddit_lies
"I like to promote intellectual curiosity, except when it comes to a blind a deaf girl who could communicate only with a special nanny and when her special nanny went away completely changed her communication style"
English
Well, I suppose if impeachment is foreclosed the traditional remedy of "firing her out of a cannon into a wood chipper along with the Kentucky supreme court" is back on the table
Lexington Herald-Leader@heraldleader
KY Supreme Court terminates impeachment of Fayette Judge Julie Goodman kentucky.com/news/politics-…
English
Man I genuinely don't get Orban
Saint Javelin@saintjavelin
Bloomberg just dropped the Orbán tapes and it reads like a bad fanfic. Hungary’s prime minister spent a phone call with Putin calling him a “lion,” casting himself as the helpful little “mouse,” offering Budapest as a venue to end the war on Russia’s terms, and closing with “I am at your service.”
English
If she's actually attractive and not a total bummer to be around then this lady is probably quite a catch, why are so many guys on here so insistently gay about this kind of thing
This is plenty young enough to have a couple of kids which is all most people want
Daniel@danielgothits
Lol it’s 2026 and women still don’t understand what makes a guy want to marry someone
English
@onefiftyfivemm @hradzka If he's not real, then how did Kevin McCallister find the hotel lobby?
English
again: *normal professional politicians just do not take wonderfully weird and memorable pictures like this*
The White House@WhiteHouse
Icons. 🐰
English
Is last samurai getting mastered and commandered
Gun Samurai@_Gun_Samurai
Are you a fan of The Last Samurai? Actually; I am. It’s fun, well made and easy to enjoy. It’s not entirely accurate and even now people still misinterpret ‘who’ the last samurai are but anyway... If you’ve not seen it, Gun Samurai recommends it. Just don’t take it seriously 🙏
English
Buddy I'm pretty anti-palestine and even I blanche at just nuking the West Bank
Iran time@khanadan99
Who wants to see Israel like this?
English