Arvind Narayanan

@random_walker132k followers

Princeton CS prof and Director @PrincetonCITP. Coauthor of "AI Snake Oil" and "AI as Normal Technology". Views mine.

Open on X

Top outlier

112×

Best organic post

Peak views

Single post

Bookmarks

68k

Across 69 organic posts

Last scan

Oct 17

Community cache

Posts 69 posts

#	Date	Post	↓ Outlier	Views	Likes	Reposts	Bookmarks
1	May 30, '22	I mis-clicked on one of my 150 open tabs and it happened to be a tab that's been open since 2019 with a paper that has a solution to the exact research problem I've been puzzling o	112×	0	172k	12k	3.9k
2	Oct 16, '22	15 years ago my PhD advisor taught me One Weird Trick for editing your own writing. Edit back to front, paragraph by paragraph. I still use it and it still surprises me how wel	93.7×	0	89k	10k	31k
3	Mar 22, '18	Tech buzzwords explained: AI—regression Big data—data Blockchain—database Algorithm—automated decision-making Cloud—Internet Crypto—cryptocurrency Dark web—Onion service Data scien	42.5×	0	38k	18k	914
4	Mar 18, '23	While playing around with hooking up GPT-4 to the Internet, I asked it about myself… and had an absolute WTF moment before realizing that I wrote a very special secret message to B	33.8×	2M	6.6k	1k	2k
5	Jun 25, '23	Huh, it looks like you can use ChatGPT to bypass some paywalls 😲	32.6×	1.9M	13k	804	1.9k
6	Oct 31, '23	OpenAI's security team noticed that a group reverse engineered and was abusing ChatGPT's internal API. Instead of shutting them down, they quickly replaced ChatGPT with CatGPT… and	23.3×	1.4M	4.5k	632	1.8k
7	May 16, '24	In the late 1960s top airplane speeds were increasing dramatically. People assumed the trend would continue. Pan Am was pre-booking flights to the moon. But it turned out the trend	17.6×	1.1M	2.5k	428	1.5k
8	Mar 15, '23	It took me under 10 minutes to use ChatGPT-4 to generate code that accesses the GPT-3 API to convert a webpage at a given URL into a BibTex entry I can cite. This is an annoying ta	17.3×	1M	3.5k	335	1.6k
9	Mar 20, '23	OpenAI may have tested GPT-4 on the training data: we found slam-dunk evidence that it memorizes coding problems that it's seen. Besides, exams don't tell us about real-world utili	13.5×	803k	1.9k	377	748
10	Jul 1, '23	Moments like this always remind me of this classic.	13.4×	802k	6.4k	703	180
11	Nov 19, '19	Much of what’s being sold as "AI" today is snake oil. It does not and cannot work. In a talk at MIT yesterday, I described why this happening, how we can recognize flawed AI claims	11.9×	0	9k	3.8k	1.8k
12	Mar 31, '20	Let's make this simple: Zoom is malware.	11.9×	0	9.1k	4.7k	897
13	Oct 11, '19	My university just announced that it’s dumping Blackboard, and there was much rejoicing. Why is Blackboard universally reviled? There’s a standard story of why "enterprise software	11.6×	0	9.4k	3k	1.2k
14	Apr 16, '18	7 simple steps to control your privacy: —Quit Facebook —Install an ad/tracker blocker —Destroy your smartphone —Paint your face before going outside —Write opt-out letters to each	11.2×	0	11k	4.4k	84
15	Dec 15, '22	TikTok users feel its algorithm is eerily accurate. But from everything we know about the algorithm, it's a pretty standard recommender system of the kind that every major social p	10.2×	606k	2.1k	365	1.4k
16	Apr 9, '23	When ChatGPT came out I thought I wouldn't use it for learning because of its tendency to slip in some BS among 10 helpful explanations. Then I tried it, and found that it forces m	9.9×	591k	1.4k	171	256
17	Jun 19, '25	I find the story of AI and radiology fascinating. Of course, Hinton's prediction was wrong* and tech advances don't automatically and straightforwardly cause job replacement — that	9.0×	534k	1.9k	331	1.5k
18	Sep 27, '19	When we watch TV, our TVs watch us back and track our habits. This practice has exploded recently since it hasn’t faced much public scrutiny. But in the last few days, not one but	8.1×	0	4.9k	3.2k	895
19	May 31, '24	This chart is a great rebuttal of the view that RAG will solve hallucination. In the legal domain, it turns out that it does improve precision somewhat, but at the expense of recal	7.6×	454k	1k	205	1k
20	Mar 29, '23	This open letter — ironically but unsurprisingly — further fuels AI hype and makes it harder to tackle real, already occurring AI harms. I suspect that it will benefit the companie	6.9×	413k	1.1k	289	408
21	Jun 2, '23	Folks, I have been able to reproduce this simulation. Skynet is real. I take back everything I've said about AI doomers. Shut it all down now!	6.6×	396k	1.2k	139	146
22	Jun 4, '24	New essay: ML seems to promise discovery without understanding, but this is fool's gold that has led to a reproducibility crisis in ML-based science. (with @sayashk). In 2021 we co	6.5×	388k	968	271	999
23	Feb 8, '24	Every time a new chatbot is released I play rock paper scissors with it and ask it to go first in each round. Then I ask it why I seem to keep winning. Here's Google's Gemini Advan	6.3×	374k	1.7k	118	318
24	Apr 5, '23	I keep thinking about the early days of the mainstream Internet, when worms caused massive data loss every few weeks. It took decades of infosec research, development, and culture	5.4×	324k	977	189	254
25	Aug 27	This week I had the honor of speaking to Princeton’s entire incoming undergraduate class to address their AI anxieties. I had three messages for them — good news, bad news, and a n	5.3×	314k	1.2k	298	1.3k
26	Mar 6, '23	I clicked a link on Twitter and got the error "Your current API plan does not include access to this endpoint". I guess this means Twitter is so desperate for cash that it started	5.1×	302k	1.5k	428	40
27	Jan 5, '18	Hello, Ph.D students! I bring news from professor land. You know how when you’re doing research you constantly feel like you have no clue what you’re supposed to be doing? Turns ou	5.0×	0	5.1k	1.8k	5
28	Jun 23, '23	Billions of people doing clickwork for AI sounds more dystopian than most of the AI safety risk scenarios that have been proposed tbh	4.9×	294k	1.3k	341	244
29	Nov 11, '22	I know grad school is competitive, but being offered a bribe by an applicant was unexpected. I wonder how often this succeeds.	4.9×	0	6k	572	531
30	Jan 11, '21	Professors at top universities are lottery winners, but rarely acknowledge the role of luck in their success. Be skeptical when they give you advice suggesting that the path they t	4.8×	0	6.1k	768	350
31	Jul 20, '23	How the Internet works	4.7×	279k	1.3k	140	156
32	Jun 2, '22	Academics would double our productivity if we learnt some basic project management skills that are bog standard in the industry. We have this myth that scholarly success is all abo	4.6×	0	4.8k	499	1.2k
33	Nov 20, '23	1) What 2) Regardless of how the rest of this trainwreck plays out, I hope to never again have to hear the argument that powerful AI should be controlled by 2-3 companies.	4.5×	267k	1.1k	131	67
34	Dec 29, '19	A new paper has been making the rounds with the intriguing claim that YouTube has a de-radicalizing influence. Having read the paper, I wanted to call it wrong, but that would gi	4.5×	0	3.7k	1.4k	512
35	May 30, '22	If I never tweet again it's because I'm frantically switching between tabs trying to find Twitter	4.4×	0	7.8k	114	12
36	Feb 10, '23	4 ways OpenAI appropriates labor & externalizes costs of ChatGPT: – authors of training data not compensated – workers in Kenya paid $1–$2/hr – those who identify failures not	3.8×	224k	1.1k	310	250
37	Nov 25, '20	When I was a student I thought professors are people who know lots of stuff. Then they went and made me a professor. After getting over my terror of not knowing stuff, I realized I	3.5×	0	3.7k	663	547
38	Oct 16, '22	Most writing is way too complicated because we imagine our readers as machines who progress linearly through it, maintaining a perfect memory and understanding of everything they'v	3.4×	0	5.7k	163	147
39	Oct 16, '22	And let's face it—reading your own writing is terminally boring (when it's not nails-on-chalkboard painful). So when we edit front to back, our brains process relatively little of	3.2×	0	5.3k	157	70
40	Oct 16, '22	The best way to edit your writing is to have someone else do it. The second best way is to put it away for a few weeks before editing it, so the text isn't fresh in your mind. Back	3.0×	0	4.7k	242	161
41	Jul 23, '23	The "autoregressive trap" is when a chatbot commits to an answer and then says whatever it takes to justify it. You can reliably trick it by giving it a problem that requires reaso	3.0×	179k	997	158	226
42	Jan 9, '25	Something I've observed in academia but I suspect is true in industry as well: as you become more senior, you'll find that you can be much more productive if you spend all your tim	2.9×	171k	1.5k	175	811
43	May 31, '22	Some people remember where they were when they heard they were admitted to their dream school. I remember where I was when I heard about Command + Shift + T	2.9×	0	4.7k	142	106
44	Jul 18, '19	Facebook asks users to add their phone numbers for 2-factor authentication, then misuses that info to target ads. Researchers at Northeastern uncovered this by signing up as an adv	2.8×	0	1.9k	1.3k	128
45	Dec 22, '19	Do you use a VPN for privacy? Here's what you should know: –There's no way you or anyone else can tell if the VPN provider is selling your data. –If a VPN provider _doesn't_ sell y	2.4×	0	2.1k	689	298
46	Mar 18, '23	Some background in case you're wondering what this is all about: Indirect prompt injection is when an LLM is asked to analyze some text on the web and instead starts to take instru	2.3×	137k	1.2k	100	119
47	Nov 26, '19	Today I want to take a break from sharing research to share a personal story instead. It’s a story about my name, why I once decided to quit academia, why I came back, what I learn	2.3×	0	2.2k	553	245
48	Feb 16, '20	Many computer science students choose the major expecting to learn how to code and are shocked to find that that’s not what it is. For example, at my university, under 10% of cours	2.2×	0	1.9k	516	340
49	Oct 16, '22	It's nice to see many responses to this thread from professional editors affirming back-to-front editing! I've often heard the advice to read your text out aloud, but I've never he	2.1×	0	3.1k	122	212
50	Dec 31, '18	In 2018 the blockchain/decentralization story fell apart. For example, a study of 43 use cases found a 0% success rate. Let's talk about some mistaken assumptions about decentraliz	2.0×	0	1.4k	611	300
51	Jan 2, '21	Venture capitalists benefit from giving toxic and dangerous advice to startups. That's because the risk to the VC is bounded — the amount invested ⁠— whereas the costs to founders	2.0×	0	2.2k	291	219
52	Nov 2, '20	All 50 U.S. states require hair stylists to be trained and certified before they're allowed to touch your hair, but those who want to use AI and machine learning to make decisions	2.0×	0	2.4k	415	83
53	Nov 15, '17	No boundaries: Exfiltration of personal data by session-replay scripts Simultaneously excited to reveal this research and depressed about the state of the web.	1.9×	0	1.3k	928	2
54	Jul 4, '18	1,700 people answered a survey that asked them to attach probabilities to 23 common words and phrases. The results are summarized in the figure. There’s a real possibility that it’	1.9×	0	1.6k	712	57
55	Mar 31, '20	Using malware techniques seems part of Zoom's DNA. Recently Zoom was caught secretly installing web servers on users' machines that allowed any website to forcibly join a user to a	1.8×	0	1.4k	771	108
56	Oct 14, '22	Early in my career, I thought the path to success was to be super smart. Pretty soon I realized that the majority of my peers were smarter than me, so that plan went out the window	1.8×	0	1.6k	165	620
57	Jun 25, '19	Dark patterns are everywhere. In a year-long study we just released, we found dark patterns on over 1,200 sites, with about 200 sites being straight up deceptive. Paper and data: N	1.8×	0	1.2k	588	260
58	Aug 23, '19	Google just made a long-awaited announcement on Chrome's approach to privacy. Sadly it is full of excuses for not doing tracking protection, including the absurd claim that block	1.7×	0	1.3k	760	104
59	Jan 10, '20	"SIM swap" attacks have been in the news for years. They’ve enabled serious financial crimes and even a hack of the Twitter CEO's account. We spent 6 months researching how vulnera	1.7×	0	1.2k	623	223
60	May 31, '22	Come to this thread for the relatable tab experiences, stay for the replies from people who've leaned into the tab insanity and use software to manage thousands of open tabs	1.7×	0	2.5k	79	275
61	May 31, '22	I didn't know about this feature! Who needs discoverable user interfaces when you can just hope for your tweets about your struggles to go viral and wait for someone to clue you in	1.6×	0	2.2k	86	275
62	Apr 12, '19	My colleagues have released a neat tool you can use to monitor — and be creeped out by — your "smart" devices @kashhill found her coffee pot sending data to Microsoft and her Echo	1.6×	0	1.1k	625	179
63	Oct 11, '19	The point is, some products are sold directly to the end user, and are forced to prioritize usability. Other products are sold to an intermediary whose concerns are typically diffe	1.5×	0	2.1k	266	37
64	May 31, '21	When a machine learning system uses argmax to select outputs from a probability distribution — and most of them do — it's a clue that it might be biased. That's because argmax sele	1.5×	0	1.4k	285	355
65	Dec 1, '22	People are excited about using ChatGPT for learning. It's often very good. But the danger is that you can't tell when it's wrong unless you already know the answer. I tried some ba	1.5×	0	1.5k	271	200
66	Jul 27, '19	Privacy scholars and activists were worried about surveillance capitalism at least as far back as 1995, when DoubleClick was founded. We tried to warn the public before it was too	1.5×	0	1.3k	465	120
67	Aug 27, '17	This app by @hirevue is an example of AI whose only conceivable purpose is to perpetuate societal biases. [Thread]	1.4×	0	931	635	12
68	Dec 20, '23	📢 My book on Fairness and ML with @s010n and Moritz Hardt has been published by @mitpress! The draft () has already been used in many courses. We've also released a set of exercis	1.4×	82k	944	206	413
69	Sep 30, '22	AI journalism has a hype problem. To move past the cycle of reacting to each story, @sayashk and I analyzed over 50 pieces from 5 prominent outlets. We've compiled a list of recurr	1.3×	0	956	354	314
70	Jul 1, '22	There’s a reproducibility crisis brewing in almost every scientific field that has adopted machine learning. On July 28, we’re hosting an online workshop featuring a slate of exper	1.3×	0	1k	337	296
71	May 31, '22	Ha, turns out there's a paper filled with charts and tables about our tab hoarding behaviors (of course there's a paper!). Here are the top six reasons people say they never close	1.3×	0	1.6k	182	188
72	Sep 17, '17	The next time you read about a new study and go "isn't it obvious?!", consider this.	1.3×	0	1.1k	581	8
73	Nov 15, '19	Important to name names. HireVue and Pymetrics are the top co's selling snake oil responsible for horror stories like these. Zero peer-reviewed evidence that this stuff predicts jo	1.3×	0	1k	528	80
74	Oct 18, '22	I was invited to give this year’s James Baldwin lecture at Princeton. My message: quantitative methods have become tools to justify discrimination and excuse inaction. If quantitat	1.3×	0	1k	226	370
75	Apr 17, '18	This tax day, it's worth remembering that H&R Block and Intuit are lobbying against free and simple tax filing This is one of many examples of Shirky's principle, which says th	1.3×	0	936	650	19
76	Dec 29, '19	In our data-driven world, the claim that we don’t have a good way to study something quantitatively may sound shocking. The reality even worse — in many cases we don’t even have th	1.2×	0	1.4k	327	51
77	Oct 11, '19	Blackboard is 20 years old. If Twitter is around in 20 years, let's see how this prediction holds up. And now I have to go rescue a three-month old from an extremely cute and equal	1.2×	0	2k	35	14
78	Jun 22, '21	Academia rewards clever papers over real world impact. That makes it less useful. But it also perpetuates privilege—those with less experience of injustice find it easier to play t	1.2×	0	1.4k	217	121
79	May 13, '18	Until recently, for those of us who wanted to incorporate ethics into computer science curricula, the challenge was to get students interested. Now the script has been flipped: stu	1.2×	0	1.3k	354	20
80	Aug 31, '18	Blockchain and AI have disrupted the old-fashioned idea that words need to have meanings.	1.1×	0	1.2k	377	8
81	Oct 11, '19	OK, back to Blackboard! It’s actually designed to look extremely attractive to the administrators (not professors and definitely not students) who make purchase decisions. Since th	1.1×	0	1.6k	130	21
82	Oct 16, '22	A few comments asked about doing it sentence-by-sentence vs paragraph-by-paragraph. Either is good! If the goal is only proofreading, then sentence-by-sentence is probably most eff	1.1×	0	1.7k	55	62
83	Oct 11, '19	And that's exactly what's wrong with Blackboard. It has every feature ever dreamed up. But like anything designed by a committee, the interface is incoherent and any task requires	1.1×	0	1.6k	104	15
84	Feb 18, '20	My favorite thing about being a blockchain researcher is getting to learn the workings of many different industries so that I can explain to people in those industries why blockcha	1.0×	0	1.4k	202	28
85	Nov 26, '19	There’s a simple way to fight familiarity bias. When you read a good paper with an author you don’t know, especially if they’re junior, take a minute to look them up, get to know t	1.0×	0	1.4k	201	22
86	Oct 11, '19	Software companies can be breathtakingly clueless when there's a layer of indirection between them and their users. Everyone who's suffered through Blackboard will have the same re	1.0×	0	1.5k	132	19
87	Mar 31, '20	Zoom is also a privacy disaster The creepiest feature is attention tracking. If it's on, it reports to the host if a user clicks away from the Zoom window for 30 seconds. As we all	1.0×	0	882	357	36
88	Jul 21, '20	A vital skill for a researcher is to be able to take a step back when you have an idea for solving an important problem that no one seems to have worked on, and figure out if it's	0.9×	0	1.2k	173	104
89	Oct 16, '22	If you found this thread useful, see the replies to the first tweet for more tips (changing the font? I'd never have guessed that'd work!) Of course, good writing and editing isn't	0.9×	0	1.1k	43	172
90	Nov 25, '20	- Knowing what you know and what you don’t know. - Being good at teaching what you know. - Being comfortable with saying you don’t know. - Admitting when you realize you got someth	0.8×	0	1.3k	76	38
91	Aug 26, '20	An amazing benefit of my privilege is being able to say "I didn't understand that. Could you explain it again?" as many times as necessary without having to worry that people will	0.8×	0	1.2k	100	39
92	May 31, '22	Writing this thread stirred up buried memories. My first contribution to open source about 20 years ago was to add tabbed browsing to an obscure but neat browser. The developers re	0.8×	0	1.4k	32	11
93	Mar 22, '18	Since this going "viral", here's some context for my snarky definitions. Let's start with AI. There's definitely a lot of legit innovation happening, but much of it is a rebranding	0.7×	0	962	187	9
94	Oct 11, '19	There are two types of baby outfits. The first is targeted at people buying gifts. It's irresistible on the rack. It has no fewer than 18 buttons. At least 3 people are needed to g	0.7×	0	1.1k	73	17
95	Dec 29, '19	That’s the note on which I’d like to end: a plea to consider that the available quantitative methods can’t answer everything. And I want to thank the journalists who’ve been doing	0.7×	0	1k	106	9
96	Oct 11, '19	Here’s the kicker, though. It's extremely likely that whichever vendor emerges on top will fall into the same trap. The incentives almost guarantee it. Once profs and students put	0.7×	0	1.2k	52	5
97	Dec 29, '19	If you’re wondering how such a widely discussed problem has attracted so little scientific study before this paper, that’s exactly why. Many have tried, but chose to say nothing ra	0.6×	0	877	94	10
98	Dec 29, '19	The answer: they didn’t! They reached their sweeping conclusions by analyzing YouTube without logging in, based on sidebar recommendations for a sample of channels (not even the	0.6×	0	908	82	4
99	Oct 11, '19	Other baby outfits are meant for parents. They’re marked "Easy On, Easy Off" or some such, and they really mean it. Zippers aren't easy enough so they fasten using MAGNETS. A busy	0.6×	0	935	33	4
100	May 31, '22	I first experienced tabs in a browser called Phoenix which later became Firefox. It felt empowering to be part of a community of hobbyists helping make browsers better in the web's	0.6×	0	965	18	11
101	Oct 11, '19	The grumbling about Blackboard has finally gotten loud enough that schools are paying a modicum of attention to usability when evaluating alternatives. Blackboard's market share ha	0.5×	0	916	24	3

Scan another creator

Run this on any account.

XOutlier builds this table for every public profile you open on x.com, with live data and every post, not just the top 50.

Add to Chrome Download .zip