I’m not like the other Bayesians. I’m different. Thinks about philosophy of science, AI ethics, machine learning, models, & metascience. postdoc @ Princeton.
Philadelphia
- Tweets 897
- Following 8,779
- Followers 32,013
- Likes 77,581
Pinned Tweet
My article on AI for science, in which I characterize a deviant notion of scientific objectivity rooted in the impossible ideal of theory-free inference, is available now open-access in Erkenntnis link.springer.com/article/10…
7
49
275
39,646
This is junk science that could be extremely harmful to vulnerable populations if ever deployed. AI facial analysis cannot infer latent psychological, personality, or behavioral features, and technology that pretends to have these capabilities is grossly unethical.
“Data like this is incredibly rare for AI systems." Learn how #USF researchers are using AI to help diagnose childhood PTSD in @FuturityNews.
Read more 🔽
futurity.org/facial-expressi…
11
262
1,410
40,956
nothing like getting a PhD in some extremely niche subject only for it to become, overnight, the one thing every human being has a clinically bad opinion on.
14
42
1,160
50,450
Chasing an academic career is optimizing for the narrowest imaginable market that is liable to simply implode at any given time and has no transferable skills or accolades.
17
61
1,146
80,852
Hello friends, a life update is in order. I am now Dr. Andrews. I will spend the next two years at Princeton as a postdoctoral fellow. I am, however, still on the job market for something permanent.
50
13
875
45,361
Princeton University policy mandates that you specify an emergency contact from the beyond
13
28
738
26,259
Heartbreaking: this postdoc did not get the job they applied to several years ago and have no recollection of having applied to
11
2
695
38,228
can’t believe I went to the trouble of getting a PhD and didn’t even get a sword out of it.
17
24
654
23,472
What most offends me about liberalism is the pretense that systems and persons that explicitly value capital accumulation over human life can be politically moved on humanistic grounds. Appeals to humanity have no political efficacy here.
11
49
354
14,414
I wish that everyone in the machine learning community understood that investing in, diverting their attention to, and parroting the hype on LLMs is the best possible way to ensure another AI winter and the complete and total irrelevance of their work in 5 years.
11
31
342
13,064
The only reason to refer to technology as “AI” is to confuse people. Change my mind.
18
18
325
64,132
Only a “paradox” when you ignore the human toll, the environmental toll, and the fact that the technology manifestly doesn’t do what it’s marketed as doing.
Research reveals a paradox: the more people understand AI, the more hesitant they may become to embrace it. s.hbr.org/46DQLOW
6
51
316
14,665
This is going to come as an absolutely shock to many of you but it has been hypothesized that language evolved as a means of communication
The average complaint about Duolingo is that it's addictive, fun, and doesn't work. But why does it not work? Humans are made to learn language. It seems like if you have gamification, the language learning should happen if only by accident.
6
5
252
14,771
My thoughts: LLMs are very useful for a very limited range of use cases. They are deceptively marketed & only apparently useful in the majority of applications in which they are currently deployed. In many of these applications, they are apt to have catastrophic repercussions.
9
21
225
11,039
The vision of human-out-of-the-loop science fundamentally mistakes what knowledge production is. It aims at the production of, wait for it, knowledge: a thing that definitionally involves an epistemic agent or agents, unless you are theistic about it.
We need new rules for publishing AI-generated research. The teams developing automated AI scientists have customarily submitted their papers to standard refereed venues (journals and conferences) and to arXiv. Often, acceptance has been treated as the dependent variable. 1/
7
27
171
19,664
Thrilled that my paper on machine learning in science and the myth of theory-free inference has been accepted for publication in Erkenntnis. The latest version of the manuscript is available for download as a preprint through the archives. philsci-archive.pitt.edu/260…
4
16
170
7,290
AI usage accounted for less than 1% of online activity and “the most consistent predictors of AI use across studies were…Machiavellianism, narcissism, psychopathy,” i.e., dark triad personality traits. Nice. liebertpub.com/doi/10.1177/2…
7
27
143
9,459
Methods in the Philosophy of Science: A User's Guide is now a real, live book. My chapter offers an appraisal of philosophy of science in its engagement with applied mathematics. mitpress.mit.edu/97802625522…
5
21
136
14,669
women changing their names after marriage is a deepstate plot to muddy their citation records
3
3
127
4,555
Thrilled to be at UNAM this week and to be keynoting a truly fantastic conference on the Philosophy of Computing.
6
1
128
9,996
python scripting has always felt like talking baby talk at a computer until it nods pityingly and goes and executes something in C
3
7
119
4,895
I am thrilled to have taken a position with the @PrincetonCITP and the @PrincetonAInews. I am still on the hunt for something permanent. If you hear of any tenure-track positions in AI ethics, philosophy, STS, information & data science or similar, please let me know.
4
6
117
6,500
2025. An era of smart machines. The vending machine eats your $3 and reports that every soda you can see is SOLD OUT. Your washing machine inexplicably sends usage data to a manufacturer of autonomous weapons. Your husband is cheating on you with autocorrect.
2
21
110
4,629
Is it an absurd privilege to be granted the right to spend all of my time answering the questions that I find most interesting? Yes. Do I think I am good at it? Yes. Will I continue at it? Yes. But its methods of self-perpetuation & accreditation are in desperate need of reform.
4
2
114
5,727
the best kind of joke is the kind where the approximately 7 people with the cultural references to get it are not stupid enough to find it funny
4
2
108
5,264
PTSD is a cognitive disorder, not a disorder of facial morphology or expression. Facial morphology and expression will never provide a reliable inferential basis for a brain condition.
2
9
108
2,563
OpenAI Admits Thing We Have All Known Since The Getgo
OpenAI admits AI hallucinations are mathematically inevitable, not just engineering flaws computerworld.com/article/40… #cwenglish #feedly
3
21
106
5,537
In place of the lofty pursuit of producing knowledge we were promised we’re trapped playing 3-dimensional chess against some of the most regressive institutions in existence.
1
2
109
4,765
Woops, you optimized for journal articles over books? You’re worthless on the UK market. Woops, you optimized for coauthored work over solo-authorship? You’re worthless on the philosophy job market. Did the reverse? You’re worthless in CS and adjacent fields.
1
2
106
4,382
The level of brightness standardized in modern headlights is a known safety hazard. Regulators and insurance safety associations think the tests circa the 1980s don’t need updating. Auto manufacturers have perverse incentives and have found ways to cheat these tests.
2
9
94
6,285
it’s so hot how departments that are not philosophy departments will disclose teaching load in adverts for faculty posts so you don’t spend six months applying and interviewing and flying out for a job that’s secretly a 5/5
1
5
92
7,440
Setting aside the casual bigotry of gatekeeping the aesthetics of queerness…
Queer aesthetic, so-called “androgyny,” is an easily spoofed signal. Any narrow-minded, misogynist creep can get a little earring and paint their nails. Queerness-as-praxis, queerness as a mode
Tying non-binary to presentation is…
I’m never gonna give androgynous. I’m 6’4”, 200lbs, and got my daddy’s entire face. You take one look at me and you gonna know exactly what’s on my birth certificate.
I’m still nonbinary because the binary structure and socialization of…
2
4
84
6,847
my favorite scifi author has liked my tweet. I will be riding this high for the next five years and will not be taking questions during this time out of respect for the family. thank you.
1
1
81
3,454
If facial morphology/pose/expression were added to the diagnostic criteria for a cognitive disorder it would be deemed laughable. The expression data these researchers used was, as far as I can tell, not even stim-locked.
2
69
1,940
What a fascinating state of play. Thousands of people commenting on this video are variously certain that it is real footage or that it is AI-generated. Their justifications for these claims are almost entirely irrelevant non-explanations.
so is this the first time AI is popping off undetected for just making up a fake crime video? no one in the comments mentioning it’s obviously AI???
Community note
CBS News confirms that the footage is real, taken from a camera outside of Al-Amira Jewelry in Anaheim. KCAL news also covered the story and showed both the viral footage and video from inside the store during the attempted robbery.
cbsnews.com/losangeles/new…
piped.video/watch?v=0Z2cah
1
15
68
11,523
Update on Methods in the Philosophy of Science: A User’s Guide. The book will become open access in July of next year. If you would like to read my contribution before that date, please feel free to message and I can send you a PDF.
Methods in the Philosophy of Science: A User's Guide is now a real, live book. My chapter offers an appraisal of philosophy of science in its engagement with applied mathematics. mitpress.mit.edu/97802625522…
1
4
63
4,816
Spotify has begun to funnel its revenue towards “AI powered” weapons tech. If you do not want to be bankrolling that, consider quitting the platform. There are plenty of good alternatives out there.
5
12
60
4,551
Please let me know if you hear of opportunities for which you think I would be a good fit: tenure-track lines or lectureships in philosophy of science and technology or applied ethics, STS, responsible AI, data science institutes, information schools or similar.
2
4
66
5,401
Many of you have, for many years, asked me to create a substack. Here I humbly submit my first offering: the one thing you least want to hear about. open.substack.com/pub/thecyb…
5
3
61
7,459
You want a ceasefire? You want aid let in to Gaza? Mass, general strike. You want environmental protections? Mass, general strike. The systems we are engaged with here only speak one language, and it is the language of capital.
1
3
59
977
This market is a soap bubble and we all know it. Claims of the capacity to obviate or automate human labor are absurd. We know this. But that won’t stop them from disrupting labor markets and, with them, human lives. pluralistic.net/2025/09/27/e…
5
11
65
4,083
Can’t wait to work at a fancy school that pays for digital access to at least one (1) journal in my field.
2
2
63
3,403
Providing relief to families in Gaza is the most urgent and effective act of altruism you could take right now.
Remember the sweet boy with a big smile. The Sameer Project team in the South got him and his family a tent and mattresses thanks to your generous support. Donate here: chuffed.org/project/113222-t…
Khan Younis is now no longer accessible to the Sameer Project team due to the displacement map release today dividing the South from the Central area. This is disastrous as around half a million people are trapped in Khan Younis.
For the time being, the Sameer Project will set up tents in the Central area where a lot of displaced people are with no shelter or tents in bad condition. Each tent costs $384 with 7% commission.
Please donate to the South campaign for more tent distributions: chuffed.org/project/113222-t…
Other ways to donate include:
paypal.me/mahertali (Paypal option, please make sure to add a message saying "South aid")
account.venmo.com/u/Maher-Al… (Venmo option, please make sure to add a message saying "South aid")
3
8
59
4,123
I have a paper with Andy Smart and Abeba Birhane on this: sciencedirect.com/science/ar…
2
8
60
1,819
One day in the future—a day that will appear to come all too soon—the headache-inducing deep learning sloppification of digital photography will have become the much-emulated nostalgic media artifact of the present era, the tape warble of our generation.
5
6
61
3,843
OpenAI Admits: Humans are Mammals
OpenAI admits AI hallucinations are mathematically inevitable, not just engineering flaws computerworld.com/article/40… #cwenglish #feedly
2
60
3,478
A word from the wise: not only can you navigate to “privacy and safety” settings on this app and revoke the right to use your content as training data, you can also block “grok” and mute any mention of the word “grok” to make your experience here marginally less cursed.
1
7
56
2,902
in some indeterminate number of weeks it will, as though by magic, become public property. until then I can email it.
10
2
58
1,391
You can waste your time bemoaning this or you can realize in it a political opportunity: there are no political agendas too lofty to be achieved if we are willing and able to coordinate and withhold our labor.
1
1
53
2,034
Feel like there was some reason we didn’t learn about the heroic citizens who hid Jews and helped them flee from the Nazis while their countries were still Nazi occupied. Can’t quite remember it now. Maybe it’ll come back to me.
Have any Israelis snuck starving Palestinian children out of Gaza, through Israel’s illegal blockade, past apartheid checkpoints to hide them in their attics to feed them? To save them?
There must be heroes, somewhere. Where are they?
3
1
55
4,495
thank you, thank you *bows deeply* i had to steal someone’s tex template and hack it myself
2
50
2,821
99 out of 100 things I see people using LLM-based chatbots for are things that I have to believe people wouldn’t be doing with them if they had an elementary understanding of the technology. Those 1-in-100 cases can be excellent, though.
2
1
51
2,430
Important point, but I think that there are a lot of critical failures of analogy here. The reliability of Wikipedia lacked a strong intrinsic upper-bound. There were not billions vested in getting students hooked on a falsely-advertised and misapplied technology.
I was born in 1993 and when I was in middle school we had lessons on how to do research on the internet. We got told that nothing on Wikipedia was reliable. We eventually learned from experience that that was false, and it burned trust w/ teachers about such things
1
6
50
5,228
Finally, someone willing to speak some truth:
A personal update.
bhaskar-mitra.github.io/post…
1
10
50
5,206
Anyone claiming this dynamic is field-specific has never been on the academic job market. The failure of mutual intelligibility between, e.g., US and UK markets is pandisciplinary.
3
1
52
3,813
ML based facial recognition is not reliable in the wild; like many “AI” tools sold to law enforcement, it is thinly-veiled pseudoscience and will harm people by lending law enforcement both a felt sense of epistemic authority and legal impunity. techpolicy.press/ices-reckle…
2
25
50
5,391
The Human Rights Council of the United Nations: The genocide in Gaza persists because it is profitable. The special commission has meticulously documented the corporations heavily invested in the continuation of this human rights atrocity.
1
15
47
4,342
I also received this wonderful moleskine notebook and a bag to keep it in from the conference.
1
47
3,199
Want to resist automation? Don’t want to be forced to work with AI? Find out just how useful your company thinks AI is when none of its workers show up one day.
1
38
1,652
It is a repugnant affront to Jews and to Jewishness that the German state requires one to affirm Israel’s right to exist to gain access to German citizenship.
Many of you have, for many years, asked me to create a substack. Here I humbly submit my first offering: the one thing you least want to hear about. open.substack.com/pub/thecyb…
1
4
40
3,110
I feel that philosophers use the Latin “qui bono” to mean the equivalent of the modern english colloquial “what gives.” Is this legitimate usage? Is this usage intelligible to the general populace (of academic philosophers)? Is there a better Latin idiom for this sentiment?
9
3
38
6,518
Yet another instance in which we see the glaring need for descriptive, useful terminology, of which “AI” is not. At this point, insisting on referring to technology as “AI” is just deliberate obfuscation. nature.com/articles/d41586-0…
2
6
35
2,675
OpenAI admits airplanes not capable of flight underwater
OpenAI admits AI hallucinations are mathematically inevitable, not just engineering flaws computerworld.com/article/40… #cwenglish #feedly
1
8
37
2,553
For my American followers: you have until the end of the year to make home improvements in the name of energy efficiency or purchase a green vehicle before green energy tax credits are eliminated.
1
6
36
2,714
“Man will not learn language until that language is all that stands between the man and pissing himself” — ancient proverb
1
1
36
1,752
read incandescence if you like to think about how the discovery of mechanics would have gone if Sir Isaac Newton were a honeybee in a world that is gravitationally and topologically radically unlike our own
4
37
1,911
( to be clear, it would still not be a “paradox,” because that’s not what that word means )
31
1,365
starting photojournalistic project where I take pictures of strangers powerpoint slides about AI that I see on the train. tune in for high quality content.
2
35
2,144
If our ancestors on the African savannah had needed button-pressing for survival maybe language would have evolved for that and duolingo would work, idk
1
32
1,069
“I can learn language from an audiobook” “I can learn language from an app” — guy who has zero conception of what a language is
2
28
1,933
There is a lot of fuss today over whether chatbots can replace human participants in social sciences research when the solution is obvious: ask chatbots to simulate the views of social scientists and survey them on attitudes towards chatbots as substitutes for human subjects.
1
3
29
1,813
clergy in the middle ages anticipated google scholar. another deepstate plot they won’t teach you in history class.
2
27
1,390
