Teaching AI. Previously Director of AI for NYC.
NYC
- Tweets 30,323
- Following 1,033
- Followers 5,802
- Likes 38,548
This is a letter Feynman wrote to a former student who wrote congratulating him for the Nobel. I’ve posted it before but I really find it worth it to read especially as a student or early stage research person.
I was recently on a panel with several other professors and we were asked to give some tips to graduate students in machine learning. It got me thinking about why professors are so bad at giving advice. So here are some reasons why you should not take advice from professors.
41
1,256
5,748
1,365,505
This is so obnoxious and really just indicates how ignorant this poster is about how anything works. This is a PhD student. Not the secretary from The Devil Wears Prada.
15
123
3,764
115,379
Honestly if I couldn’t convince with sources I’d probably just resort to credentialism too.
8
7
3,101
62,125
I used to work in Fischer Black’s old group at Goldman Sachs; his colleague Bob Litterman was the chairman of the group then. There was one episode where Fischer was confused about some options’ market price and realized people were misusing B-S and cleaned everyone out.
Fischer Black, after inventing Black-Scholes, eventually quits academia and goes to Goldman Sachs, and literally just doesn't care about the money, just wants to hang out and solve interesting problems
15
107
2,683
493,271
No, there isn’t. This person just does not exist. You can get into college if you want, period. You could also take a random IIT Bombay admit that couldn’t get an H1B and they would absolutely decimate the kind of American kid you’re talking about, on both merit and work ethic.
There's a 20 year old American kid working at Dunkin' Donuts high everyday who is smarter than every South Asian admitted to the country in the last four years who was blocked out of college in the name of diversity.
29
45
1,503
162,649
Just drop out. Dropping out of a PhD is not like dropping out of high school. You don’t have to finish it and in fact should not if it’s not working for you. Nobody cares. It’s hard and not for everyone.
I think I will conclude that I regret starting a PhD.
Doing research was a hobby and it was unwise to turn it into a career. It's a deeply personal endeavor that should be done for its own sake.
The hobby is gone, and also my pace of career progression. Lose-lose situation.
9
45
786
85,545
This site has become kind of hard to use. I basically stay on it because academics and math people haven’t left, and that conversation is civilized and ok, but you step one toe outside that and it’s just deranged racist morons and trash ads everywhere. When are we going to move?
61
29
693
46,058
Matlab is maligned, sometimes for good reason, sometimes unfairly. But there is this thing in it called the backslash operator. A\b “solves” Ax = b. But this operator has like half of numerical linear algebra in it. It actually does this.
16
60
731
114,640
This is also more or less what it would look like if universities tried to break graduate student strikes by deploying faculty. :)
7
7
568
There is an important story from when John Nash was at Princeton. He was getting somewhere and his advisor goes, this is interesting, go show it to von Neumann and see what he says.
“What?”
There’s a similar story in “Monster Minds” in Feynman’s autobiography. He was warned but Wigner, who was the chair, stopped Feynman in the hall and said he had personally invited Pauli and John von Neumann. “I don’t know if he can come but I invited Professor Einstein too.”
3
27
692
109,079
You don’t live in the city, Andrew. Go back to your suburbs. Shut the fuck up. Your entire admin was shitting on the city and now you want this job? No means no.
2
656
3,279
My favorite one might be when he wrote to Ed Colligan at Palm: “When you say that we’re ‘both just going to waste money on a lot of lawyers,’ I feel you are overlooking the asymmetry in the financial resources of our respective companies.”
1
17
553
61,802
One of the most shocking things at the Stanford CS admit weekend is that Donald Knuth still comes. The faculty are supposed to introduce themselves:
“I’m Don and I write books about computer programming.”
8
38
572
112,835
I am highly disinclined to crap on a student. But this is a bunch of meaningless stuff. “Sailed oceans and hiked” who gives a shit. A 4.0 is meaningless too — from where? Doing what? My roommate was valedictorian at 15 in Sandusky, OH and got his ass handed to him at Penn.
12
11
593
46,598
A TON of people including Barack Fucking Obama told Ginsburg she should retire and get off the court but nobody can make them do it
1
1
549
3,621
He said an 88 year old congressperson should retire because she is experiencing clear decline and can’t do her job.
nytimes.com/2025/06/11/us/po…
3
2
543
8,227
His DNA was still very noticeable there. One of the most underrated things about Fischer Black was his writing, which was incredibly plain and clear. He learned from these Rudolf Flesch books like The Art of Readable Writing. Black-Scholes is written at like a 6th grade level.
3
9
532
20,590
If these people want to cover this kind of math/CS area, and they do need to because AI is not going away, they really need to learn more about all this to avoid sounding like this much of a rube. It’s like the Large Hadron Collider people interviewing you over inclined planes
1
2
490
14,428
There’s a story of Varadhan’s defense. Some old guy he didn’t know was sitting in the back and started asking very difficult questions and he was getting annoyed. After, he asks his advisor, who the hell was that?
“Oh, I forgot to tell you I invited him. That was Kolmogorov.”
This is what I’m gonna do if someone asks me a question I don’t know the answer to at my dissertation defense
4
28
474
61,927
This meme has been going on for a while but I never said anything because my advisor once asked a brutal question I can’t top, about some algorithm, not even to an enemy. But ok, here it is.
“Let me rephrase. Does there exist any question for which this provides the answer?”
5
21
435
58,013
I would also have been incredulous for the frustrated parent to eventually reveal himself to be Feynman
1
374
28,417
There's an important point when you have to switch from just consuming more knowledge to actually producing things. Diligent people/students often get sucked into just keep reading more books because they feel bad they don't know everything. (I'm not subtweeting anyone here.)
8
32
377
35,010
It is worth reading his originals even if you’re not in finance. Black-Scholes, “The Pricing of Options and Corporate Liabilities”, is a masterpiece.
1
6
392
18,968
For anyone responding to this by saying it’s just measures or whatever, this is a piece by a Fields Medalist also struggling with this. It is not an easy question.
eudml.org/doc/289648
8
26
353
47,014
I’m not going to crap on it fully again. But I cannot properly express the profound and unique unhappiness and loneliness I felt in San Francisco specifically. My point is if you don’t like it it’s ok. Live where you like! To me it was just horrible there.
18
6
362
60,025
Feynman also used to do this and people would try to avoid having him on the committee. Though the worst horror story I heard was when Tibshirani pointed out in a defense that some bio student did cross-validation wrong, insisted he re-do it, and all the results vanished. 😱
fun story:
terry tao was on both my and my brother's committee.
he solved both our dissertation problems before we were done talking, each of us got "wouldn't it have been easier to...outline of entire proof" 🫠
12
24
357
187,426
The first paper I wrote in grad school (12 years ago) just crossed 20,000 citations. To mark this satisfying but ultimately irrelevant milestone, I thought I’d share some thoughts on how to try to increase probability of more citations, especially since grad students follow me.
5
33
333
121,437
Let me tell you a real story about this I went through at Stanford PhD. It involves an actual magician, Persi Diaconis. So I got admitted but in PhD just getting admitted is only the beginning, you need an advisor. I wanted to meet lots of people for various reasons.
Good example of how unspoken norms can serve to gatekeep academia.
The way I've heard it put from an 'academic thoroughbred' professor was "I still had to climb the mountain, but there was a net underneath & all of the stable handholds were highlighted in neon for me"
2
14
354
75,727
This is the most PhD student schedule ever. Once you realize you don’t actually need to follow the 24 hour clock things get odd
5
5
317
24,678
You said this, right? Maybe the vibe shift will include dunking your head in a toilet.
7
4
314
25,868
I would like to talk about one point made here that you can set up meetings with anyone. YES!!! You can do that, and you should. I learned this from reading about David Shaw who treated his early finance jobs “like a university.” I teach my students to cold call. Do it!
observations from my first two weeks as a Meta research intern
- research jobs are the same everywhere: no one ever asks me what I’m doing or how I’m spending my time; there’s an implicit expectation to be interested and work hard
- biking to work in the sunshine has noticeably improved my quality of life
- you can schedule a meeting with anyone and they will actually talk to you; this is an amazing and I suspect under-utilized perk of the job
- had my first good research idea in a while and it wasn’t exactly in line with the goal of my internship; i was excited about it and folks were really supportive in letting me change course
- even though meta is a big company and has lots of process like any other, they put a lot of effort into removing barriers (unnecessary meetings, forms to fill out, etc.) and just letting you do research
- no meeting wednesdays is a good rule
- no H100s at FAIR (at least not that I can get) and A100s are few and far between
- SF office is beautiful and has great views of the Bay (see pic)
- it’s funny to come from academic setting where i would bring my own lunch to Facebook where i overheard some coworkers complaining about the dryness of the (free) duck served at lunch
- Meta is a great place to work; i like it here
3
12
303
46,478
This is not how you expect a Nobel level economist to write in the Journal of Finance. But he does, which is why I’m saying you should really read him. It’s disarming.
1
4
312
15,507
For some reason people liked this, so it made me think of other old Wall Street stories. I probably shouldn’t tell some of these, but, it was a while ago. Once, this guy in my group had the model tell him to build up a position in Krispy Kreme and, well, things happened.
I used to work in Fischer Black’s old group at Goldman Sachs; his colleague Bob Litterman was the chairman of the group then. There was one episode where Fischer was confused about some options’ market price and realized people were misusing B-S and cleaned everyone out.
4
8
309
51,644
Going to wait to see if they update it again tomorrow. Don’t understand what’s going on when US DOT said they’re grounding all of them.
270
29,255
Literally the exact thing I’ve been saying. Canada should do this at scale. China knows what’s going on but it will still be hard for them to pull this off.
8
29
264
24,427
I have gotten quite bored of AI for reasons in this general vein. It’s kind of a lame field, despite all the crazy hype and money and all that, because so much time is wasted debating total nonsense with completely random people who are just “fans” of something or another.
Many AI enthusiasts have just never thought carefully about like computer science or math or logic and I just can’t really communicate with them, I guess. They really don’t understand why it makes sense to say if it makes an illegal move 5% of the time then it can’t play chess.
5
13
256
19,933
Clearly a skill issue here because they also gave away that the only reason they wanted a quote was to find one to dunk on it for being problematic in some way, not to actually just listen to whatever they were saying and report it straight
3
229
11,389
Black-Scholes was rejected :)
Mamba apparently was rejected !? (openreview.net/forum?id=AL1f…)
Honestly I don't even understand. If this gets rejected, what chance do us 🤡 s have.
4
18
252
216,436
I agree, I suspect Kamala Harris’ easy laugh and sense of humor, combined with the strategy of rightly calling Republicans total weirdos, will be effective and also not easy to run against.
New pod: Trump/MAGA sneering at Kamala Harris's laugh is a big political blunder. Her fun, optimistic demeanor is an effective antidote to MAGA's vicious, hateful politics. MAGA is too locked into "American carnage" to see this.
@jenancona and I dig in:
newrepublic.com/article/1842…
6
24
178
27,533
You live in Indiana. You do not pay your own way at all; your entire state can’t support itself and takes in federal funding from other people around the country to subsidize you.
1
2
241
2,164
When I was younger, I wanted more than anything to be one of those guys. But it is not a thing you can just choose to be, and once you get older you have to let it go. Just do your own work as best you can and remember everyone is worse than someone else they admire.
3
2
244
6,092
It’s true. I had no interest in academia when starting at Stanford. But every other profession is also busy work and junk. Most hedge funds do not deliver alpha. Nobody reads your crap in the NYT. Your startup will go bankrupt. Most jobs are a fraud if that’s your sense.
As an undergrad, I thought about going into academia, but I never could understand the publish-or-perish system.
I could see that the vast majority of articles published in my field (history) were dreck.
Why make people publish if they have nothing to say? What is wrong with having professors who primarily aim at being good teachers?
7
4
231
76,886
There is a now obscure book everybody interested in optimization should read; we call it the Red Book. The real title is Optimization by Vector Space Methods, by Luenberger. It’s really about convex duality and functional analysis. It’s lovely and how I learned duality.
6
20
220
33,660
But this paper, just called “Noise”, is published in Journal of Finance, the fanciest journal, but reads like a conversational note if you look at it.
onlinelibrary.wiley.com/doi/…
3
5
232
16,984
Forget “progressive commentators”. Where are Bill Gates and Bezos? Where are the heads of all the Silicon Valley companies that entirely dominate the US economy and stock market? Where is Jamie Dimon? The US Chamber of Commerce complained to their credit but not strongly enough.
Why is Richard Hanania defending Indian Americans more strongly than any progressive commentators I've seen??
Where are progressives in this fight??
11
15
223
39,179
You do your own work. Don’t waste time on stressing out that Feynman or Einstein is down the hall. That way lies madness and pointlessness. Even for the guy who scares you, there is another guy who scares them.
2
9
226
8,136
It is worth reading what Tao said here. It has a lot in it and it is not just “cutting research funding is bad” from a heavyweight. It has many personal details of what is being destroyed here and things about eg compressed sensing I did not know either despite being around it.
Terence Tao has written a thread on Mathstodon about the damage being caused by the grant freezes at UCLA.
mathstodon.xyz/@tao/11495684…
3
32
222
21,852
There’s a similar story in “Monster Minds” in Feynman’s autobiography. He was warned but Wigner, who was the chair, stopped Feynman in the hall and said he had personally invited Pauli and John von Neumann. “I don’t know if he can come but I invited Professor Einstein too.”
1
218
116,077
Everyone involved in the computational sciences should read this. It’s not mine. These solvers are so fast, and this is just one guy doing it for a PhD not a full commercial thing, that you have to measure it in hertz. These solvers are in SpaceX rockets.
cvxgen.com/docs/index.html
1
22
210
34,072
One time there was a conference at the Institute for Advanced Study on algebraic geometry. Peter Sarnak and Nick Katz, both former chairs of Princeton math, were presenting progress on some problem. Pierre Deligne solved it in the lunch break between the two parts.
1
1
209
7,582
I mean I totally get his frustration but actually just starting a whole separate PhD without telling the advisor is so funny that it’s hard to get mad at it.
3
7
198
20,734
The reason this is important and not just fun is because you must never get discouraged. There is always some other guy. Even for Nash, so brilliant his entire recommendation letter to Princeton just said “John Nash is 19 years old. He is a mathematical genius.”
1
207
5,814
Nobody wants you to drop it. Do more and make sure every single voter knows what you people think please. Thanks.
1
174
3,451
There was one talk I remember that involved Stein’s paradox. Some old guy in the back corrected the speaker at one point. It was Charles Stein and everyone started laughing.
1
3
208
5,865
This is the thing with Wall Street. It’s a couple things at the same time. Normal people think of investment banking which has nothing to do with this at all. But sometimes you put on a trade and it turns out goddamn Fischer Black was on the other end of it. It has culture too.
3
2
203
11,059
They seem to do this with everything. Even after the 2016 election they were still incredibly angry
5
153
12,719
You know, rural people also do this exact thing and act like people who live in cities aren’t even American.
1
1
160
5,704
This is an old line from John von Neumann. “People who do not understand mathematics is simple don’t understand how complex the world is.”
People who think agi is imminent aren’t overestimating how quickly ai will improve. They are underestimating how complex the world is. Being an expert on deep learning doesn’t necessarily make you an expert on how the world works.
5
17
180
14,280
“It’s far enough now. Go show it to Professor von Neumann and see what he says.”
Nash, correctly, is scared shitless. So he goes and tries to explain the premise and von Neumann cuts him off quickly and says, “Sure — this is the fixed point theorem, yes?” Sigh, yes. Nobel Prize.
4
200
6,203
I know people are excited by LLMs. They are miraculous despite their issues. But what I constantly have to remind people is that almost everything that runs actual really serious stuff - courts, hospitals - is linear and logistic regression, with variants. Period.
13
22
189
24,763
Statistics is really difficult to teach. I took a course in statistics at Wharton and I think I had done enough math by that point to know about Galois theory or whatever. It was all these cookie cutter hypothesis tests. I retained nothing. I actually learned it in grad school.
When people say we should replace calculus in the high school math sequence by "statistics" what they often mean is something like, "it would be good if students had some statistical literacy." But it is not clear to me that teaching about e.g. chi^2-tests would provide this!
4
5
182
45,950
From “Perfectly Reasonable Deviations from the Beaten Track: The Letters of Richard P. Feynman.”
4
9
181
28,366
Homework is good and a lot of people need more of it and harder homework
I actually can’t believe that homework is normalised. Like kids go to school Monday to Friday 8.30 to 3.30 38 weeks of the year, get home about 4/4.30 (or later if clubs) and are expected to then spend another couple of hours on school work. Why do we accept this?
6
1
180
23,281
This reminds me of this very famous slide from @JeffDean. This is from 2010 so of course the numbers will be different now. From this great presentation:
static.googleusercontent.com…
1
29
180
39,325
There are also the “Financial Notes”. I think these were in most cases internal to Goldman Sachs but also circulated a bit informally. If you can find them they were quite interesting, they’re numbered like “Financial Note No. 4”. I had read them all ages ago.
1
3
176
17,987
These sentences are very short and use simple sentence structure and the simplest vocabulary. That is from Rudolf Flesch.
2
173
11,752
I am interested in this Sanskrit thing, is there a group working on this?
I’ve left OpenAI!
Already miss everyone on the Training team & my friends ❤️ but very excited to soon announce what’s next
Until then, I’ll be taking a break to solve OCR for Sanskrit so we can immortalize the classical Indian literary canon in the weights of superintelligence
6
6
172
22,597
I hesitate to comment on Columbia but there is a ton of BS on here. Let me give a different perspective. I teach in the School of International & Public Affairs, a place with an actual Middle East Institute and a lot of political content and so on.
5
17
157
121,790
I actually think linear algebra is one of the most difficult topics in math since there are so many different angles on it and you have to be able to switch among them. Also, I did a lot of linear algebra interviews on Wall Street, and people don’t know it even if they got an A.
12
6
162
13,220
Stephen Boyd would review and send edits for my LaTeX code (even when it typeset correctly) even in my first year. This kind of thing was one of the reasons I ended up working with him even though I wasn’t really trying to do optimization.
It's always nice when very busy/successful people take time to engage in a depth (e.g., suggesting edits in docs) that less busy/successful people don't.
This is because they know details matter, what excellence looks like, and that glamorousness + impact aren't very correlated.
7
4
164
20,330
Maybe I should be clear I have zero opinion on this paper. I haven’t read it. I just don’t like people especially with big accounts dunking on students when there is nothing abnormal at all about a 3rd year PhD student being the lead author on something like that.
1
1
162
3,850
An “internship” at that level is just the HR/legal structure that’s created to let a PhD student effectively be a visiting researcher for a summer. These are not the interns who the media corps pay nothing to and just get you coffee. It’s a visiting researcher.
1
1
159
3,878
Absolute nightmare situation for literally the first talk you give as a 20 year old or whatever he was. He said Pauli asked some question he didn’t properly understand until many years later but he saw there was some problem in the theory. “Wheeler will never give that talk.”
1
160
7,504
It is a lovely letter, and always puts me in a mood. The most important part:
“Know your place in the world and evaluate yourself fairly, not in terms of the naive ideals of your own youth, nor what you erroneously imagine your teacher’s ideals are.
Best of luck and happiness.”
This is a letter Feynman wrote to a former student who wrote congratulating him for the Nobel. I’ve posted it before but I really find it worth it to read especially as a student or early stage research person.
1
32
155
24,525
“When I discussed it with John von Neumann, he had a better idea.” FFS
How Claude Shannon decided on the name "entropy" in information theory (image from the nice survey arxiv.org/abs/1405.1003):
5
14
143
21,606
This is objectively untrue, his issues barely get covered. For example, have you heard incessant coverage about his disastrous RNC speech or stuff like this link? No.
nbcnews.com/business/busines…
2
8
121
3,022
I’ll give one example. I remember when a grad student described compactness as a “finiteness condition” to me. He said it’s like a thing that isn’t necessarily finite but is “small enough” in some way to “behave” closer to finite things.
Yes. I started hanging out with math grad students early on in freshman year and started to hear how they talked about and thought about real math.
10
5
154
19,982
It wasn’t a “claim” it was a joke and not being able to tell the difference doesn’t seem like a good example of FL education
2
153
2,489
I’m really not one to get into the shitting on NJ from NYC discourse because I think it’s usually childish, but they are way way past the line here. I hope NY takes them to the cleaners.
1
125
6,664
The writer Ludwig Borne once said that the source of all genius is sincerity. I believe quite strongly this will help you find your path and partners if you really are pursuing a PhD in a serious way. Be honest and open and the ones for you will see it in you. Good luck.
5
149
4,371