okayhughes (they/them)
@okayhughes.bsky.social
Ph.D. student in Climate Science.
My work improves computational+mathematical models of atmospheric dynamics in climate models.
Interested in any work that makes dialogue between policymakers, communities, and scientists more equitable.
Free palestine 🇵🇸
My work improves computational+mathematical models of atmospheric dynamics in climate models.
Interested in any work that makes dialogue between policymakers, communities, and scientists more equitable.
Free palestine 🇵🇸
If AI companies succeed in putting everyone out of work with models trained on shit stolen from the internet, I think we’re gonna all have to finally concede that the internet was a huge mistake.
November 4, 2025 at 6:35 PM
If AI companies succeed in putting everyone out of work with models trained on shit stolen from the internet, I think we’re gonna all have to finally concede that the internet was a huge mistake.
It’s so weird to get the opportunity to be a visiting scientist at one of the most beautiful and admirable scientific organizations I’ve ever come across knowing that institution could cease to exist in a year
April 19, 2025 at 4:39 PM
It’s so weird to get the opportunity to be a visiting scientist at one of the most beautiful and admirable scientific organizations I’ve ever come across knowing that institution could cease to exist in a year
FYI: if you live in an apartment building and your wifi is extremely slow (but your wired is fast), and it remains slow after swapping access points/routers, then you should probably change what channel your wifi is on (e.g. switch from 11 to 6)
March 22, 2025 at 7:06 PM
FYI: if you live in an apartment building and your wifi is extremely slow (but your wired is fast), and it remains slow after swapping access points/routers, then you should probably change what channel your wifi is on (e.g. switch from 11 to 6)
Reposted by okayhughes (they/them)
1. Sabotage public institutions.
2. Claim they don’t work efficiently.
3. Eliminate them.
—
That’s the playbook.
2. Claim they don’t work efficiently.
3. Eliminate them.
—
That’s the playbook.
January 23, 2025 at 4:20 PM
1. Sabotage public institutions.
2. Claim they don’t work efficiently.
3. Eliminate them.
—
That’s the playbook.
2. Claim they don’t work efficiently.
3. Eliminate them.
—
That’s the playbook.
Someone at (unspecified conference that happens this time of year) just delivered a talk about “AI emulators” that was actually principle component analyses.
Huge fan of this hype cycle.
Huge fan of this hype cycle.
December 10, 2024 at 5:58 PM
Someone at (unspecified conference that happens this time of year) just delivered a talk about “AI emulators” that was actually principle component analyses.
Huge fan of this hype cycle.
Huge fan of this hype cycle.
Reposted by okayhughes (they/them)
Carceral AI is a thing to be abolished, not built better.
Proud to have helped with this.
Proud to have helped with this.
🚨 The Carceral AI report is out! 🚫 Read the recommendations of global activists and researchers to resist technologies designed to police, incarcerate, surveil, and control human beings.
carceral-ai.com
Download the pdf on SSRN: papers.ssrn.com/sol3/papers....
carceral-ai.com
Download the pdf on SSRN: papers.ssrn.com/sol3/papers....
begin | Carceral AI
carceral-ai.com
November 20, 2024 at 3:46 PM
Carceral AI is a thing to be abolished, not built better.
Proud to have helped with this.
Proud to have helped with this.
It’ll be very satisfying if the net result of AI is that basically every business that’s run by McKinsey-poisoned MBAs who only care about maximizing quarterly earnings automates half their workforce with half-baked AI and drives themselves out of business.
November 20, 2024 at 4:22 AM
It’ll be very satisfying if the net result of AI is that basically every business that’s run by McKinsey-poisoned MBAs who only care about maximizing quarterly earnings automates half their workforce with half-baked AI and drives themselves out of business.
As a scientist you’re partially responsible for what happens if your research “breaks containment” and gets used by people in ways you think are bad.
November 17, 2024 at 2:14 AM
As a scientist you’re partially responsible for what happens if your research “breaks containment” and gets used by people in ways you think are bad.
Reposted by okayhughes (they/them)
the DOGE agency will not have the authority to reallocate funds (nor would that be constitutional), it is a PR tool to create the impression that government spending is inherently wasteful and build political momentum for spending cuts
Hypothetical. I'm curious what people think. If the DOGE agency worked on the premise that half of all savings in a federal program were distributed as increased benefits to qualified recipients, would you consider the details ?
November 15, 2024 at 4:45 AM
the DOGE agency will not have the authority to reallocate funds (nor would that be constitutional), it is a PR tool to create the impression that government spending is inherently wasteful and build political momentum for spending cuts
Reposted by okayhughes (they/them)
a great thing about online meetings is that you can turn off your camera and mic and yell at other people without anybody knowing
November 15, 2024 at 5:18 PM
a great thing about online meetings is that you can turn off your camera and mic and yell at other people without anybody knowing
Reposted by okayhughes (they/them)
All faculty, not just writing faculty
“Writing faculty have both the agency and the academic freedom to examine generative AI’s dishonest training origins and conclude: There is no path to ethically teach AI skills. Not only are we allowed to say no, we ought to think deeply about the why of that no.”
Burn it down: a license for AI resistance (opinion)
Resistance is not futile, Melanie Dusseau writes.
www.insidehighered.com
November 12, 2024 at 3:12 PM
All faculty, not just writing faculty
the newest llms’ performance on olympiad problems has mostly just convinced me that olympiad-style problems are even less correlated with anything that actually matters than I already thought.
September 26, 2024 at 4:38 AM
the newest llms’ performance on olympiad problems has mostly just convinced me that olympiad-style problems are even less correlated with anything that actually matters than I already thought.
Reposted by okayhughes (they/them)
Have a discussion at the beginning of such a class about the difference between "completing assignments" and "learning," and about the purpose of assigned homework. Trust students to do a good job, and radically accept the existence of a nonzero number of assholes who will find ways to cheat.
September 25, 2024 at 6:56 PM
Have a discussion at the beginning of such a class about the difference between "completing assignments" and "learning," and about the purpose of assigned homework. Trust students to do a good job, and radically accept the existence of a nonzero number of assholes who will find ways to cheat.
the suggestions under ebay listings for lampworking reference books are like 50% extremely low-rent smut and I’m SO HERE FOR IT
September 3, 2024 at 4:40 AM
the suggestions under ebay listings for lampworking reference books are like 50% extremely low-rent smut and I’m SO HERE FOR IT
exciting news
September 2, 2024 at 1:24 AM
exciting news
ok yes in hindsight it makes sense that you would use an obscure diameter of fitting for fuel/oxygen torch fittings so you don’t get some bozo doing their >2000° plumbing with stuff from lowes
but also i, bozo, wanted to do my >2000° plumbing with shit i got at lowes
but also i, bozo, wanted to do my >2000° plumbing with shit i got at lowes
August 27, 2024 at 3:40 AM
ok yes in hindsight it makes sense that you would use an obscure diameter of fitting for fuel/oxygen torch fittings so you don’t get some bozo doing their >2000° plumbing with stuff from lowes
but also i, bozo, wanted to do my >2000° plumbing with shit i got at lowes
but also i, bozo, wanted to do my >2000° plumbing with shit i got at lowes
Reposted by okayhughes (they/them)
TW.
I have lost two friends and colleagues to suicide linked to secondary trauma. One had been taking oral histories from the war in Vietnam; the other was a Turkish sociologist working on honor killings.
‘If you don’t watch you don’t care’ isn’t just untrue, it’s harmful, potentially fatally so.
I have lost two friends and colleagues to suicide linked to secondary trauma. One had been taking oral histories from the war in Vietnam; the other was a Turkish sociologist working on honor killings.
‘If you don’t watch you don’t care’ isn’t just untrue, it’s harmful, potentially fatally so.
August 23, 2024 at 5:32 PM
TW.
I have lost two friends and colleagues to suicide linked to secondary trauma. One had been taking oral histories from the war in Vietnam; the other was a Turkish sociologist working on honor killings.
‘If you don’t watch you don’t care’ isn’t just untrue, it’s harmful, potentially fatally so.
I have lost two friends and colleagues to suicide linked to secondary trauma. One had been taking oral histories from the war in Vietnam; the other was a Turkish sociologist working on honor killings.
‘If you don’t watch you don’t care’ isn’t just untrue, it’s harmful, potentially fatally so.
Ok if anyone out there is like me and hasn’t paid attention to fountain pens over the last 6 years, chinese manufacturers are making some INCREDIBLE pens at very affordable prices with actually good QA. Some of the nicest ones are even original designs!
August 20, 2024 at 3:59 PM
Ok if anyone out there is like me and hasn’t paid attention to fountain pens over the last 6 years, chinese manufacturers are making some INCREDIBLE pens at very affordable prices with actually good QA. Some of the nicest ones are even original designs!
is there a way to get your instagram algorithm to stop feeding you content that makes you feel like you’re having keys jingled at you?
August 10, 2024 at 1:54 PM
is there a way to get your instagram algorithm to stop feeding you content that makes you feel like you’re having keys jingled at you?
I’m always worried that I’m sloppy with my compute job requests but it turns out that I’ve spent $5 in cpu compute this year. oops
August 8, 2024 at 4:41 PM
I’m always worried that I’m sloppy with my compute job requests but it turns out that I’ve spent $5 in cpu compute this year. oops
introducing a new low-order interpolation scheme that arbitrarily induces new extrema at will, just to feel something
July 26, 2024 at 8:22 PM
introducing a new low-order interpolation scheme that arbitrarily induces new extrema at will, just to feel something
Does anyone have a recommendation for which numerical method for incompressible flow has a very concise implementation?
July 24, 2024 at 1:15 AM
Does anyone have a recommendation for which numerical method for incompressible flow has a very concise implementation?