James Freitag

3.4K posts

James Freitag banner
James Freitag

James Freitag

@jimfreitag

Professor of Math at UIC.

Chicago, IL Katılım Haziran 2016
330 Takip Edilen826 Takipçiler
James Freitag
James Freitag@jimfreitag·
@AliceFromQueens @feelsdesperate Certainly, things are not functioning normally at NSF and NIH. Beyond budgets, many on the ground anecdotes confirm this. In my program area, no standard grants have been awarded this year. Normally the first grants awarded would have been awarded several months ago.
English
2
2
13
2.3K
James Freitag
James Freitag@jimfreitag·
@lreyzin I also met Bill a few times. When I moved into my last office, in a file drawer behind an old stack of exams, I found a book about the history of the Incan empire with Bill's name inside the front cover (I gave it to him at some point!).
English
0
0
1
51
Lev Reyzin
Lev Reyzin@lreyzin·
I just learned the sad news that Bill Howard, of the Curry-Howard correspondence fame, passed away a few days ago. He was 99. Bill spent his career at UIC, and I was privileged to get to know him during my early years as a faculty member.
English
1
0
11
952
James Freitag
James Freitag@jimfreitag·
@JDHamkins I would expect that to be an issue with almost any university or business reimbursing travel expenses, to be honest.
English
1
0
38
2.9K
James Freitag
James Freitag@jimfreitag·
@LocasaleLab To be clear: you’re in a field where h-induces are high and meaningful from what I understand. Math is a field where some full professors at Harvard have indices in the teens. You know this (or should?) and chose to compare yourself to Terry via the metric.
English
0
0
1
68
James Freitag
James Freitag@jimfreitag·
@LocasaleLab Ok. You said some things about Tao, mentioned his H-index and mentioned yours. People who understand anything about the different subjects you are in understood it to be ridiculous.
English
2
0
1
347
Jason Locasale
Jason Locasale@LocasaleLab·
Impressive misreading. I was not comparing myself favorably to Tao. I was criticizing his selective outrage, his activism, and his silence while universities gutted due process and academic integrity for years. He only seemed to discover institutional injustice when it affected him.
James Freitag@jimfreitag

@joelwatsonfish @LocasaleLab You are replying to a person who compared himself favorably to Terry Tao on the basis of H-index, so good luck!

English
1
0
20
4.6K
Joel Fish
Joel Fish@joelwatsonfish·
@LocasaleLab You do realize that I speak with professors who are not in my department. Most of them —*gasp*— aren’t even at my institution.
English
1
0
3
385
James Freitag
James Freitag@jimfreitag·
@rand_longevity I use AI all the time. Most of my colleagues do a bit as well. I never heard people saying this. Maybe it will happen, but I’d expect if it does, society will have been so altered that this will not be close to the top headline amongst all the changes.
English
0
0
1
118
Rand
Rand@rand_longevity·
every professor I talk to that uses AI says the college system is about to collapse
English
616
542
8.5K
1.8M
James Freitag
James Freitag@jimfreitag·
@littmath @joelwatsonfish @tamaybes Right now, we all understand annals quality papers to mean the very best produced in a given year, say. But how could it ever become easy to produce such a thing? If a model can just produce papers of today’s annals quality, won’t the quality have increased quite a lot by then?
English
1
0
1
124
Daniel Litt
Daniel Litt@littmath·
@joelwatsonfish @tamaybes Trying to measure the marginal cost of generating such papers with the model. (It’s supposed to be an approximation of the marginal cost of getting a human expert to generate an Annals paper.)
English
2
1
2
178
Tamay Besiroglu
Tamay Besiroglu@tamaybes·
I bet @littmath that in 5 years AI will be able to produce Annals-quality Number Theory papers at an inference budget at or below $100k/paper, at 3:1 odds in my favor.
Tamay Besiroglu@tamaybes

@littmath @DavidDuvenaud @davidmanheim Let's say: Frontier AI models in March of 2030 will be capable of writing 2025-Annals-level Number Theory papers (as judged by you) when inference compute is scaled up to ~$100k/paper.

English
22
14
253
56.7K
James Freitag
James Freitag@jimfreitag·
@boazbaraktcs @littmath Any one person’s point estimate might not be so useful, but hearing quite a few from many people gives a good impression of the current state of the art.
English
0
0
1
831
Boaz Barak
Boaz Barak@boazbaraktcs·
@littmath As the authors themselves say, it's not just about answering a question but knowing the right question to ask. If you are staring at a tsunami, the point estimate that you are dry is not very useful.
English
3
5
134
21K
James Freitag
James Freitag@jimfreitag·
@aryehazan Cool - sorry I don’t know the area well enough to really get an impression. Sadly none of the models has been any help on problems I’ve tried beyond sometimes helping with examples. I have found they greatly assist in references and learning things though!
English
1
0
0
34
Aryeh Kontorovich
Aryeh Kontorovich@aryehazan·
I'm reaching a point where I think that the most efficient way to conduct math research is to study LLMs
English
16
5
120
9.1K
James Freitag
James Freitag@jimfreitag·
@rieszspieces @hatley_x Though it is a bit inconvenient that your context differs when considering polynomials and their evaluations at some value… 😀
English
0
0
0
172
mostly harmless graduate student
mostly harmless graduate student@rieszspieces·
sorry but major peeve of mine is the concept of "order of operations" mattering beyond like 3rd grade; there is one way to control the order of your operations, and it's called parentheses, everything else is context-dependent and/or necessarily ambiguous
Justin Skycak@justinskycak

Applied to math, this approach produces calculus students who can't evaluate 5-3x6 without a calculator because they -- can't do mental arithmetic, -- never learned their multiplication tables, -- forgot the order of operations, -- and forgot how negative numbers work.

English
89
32
1.5K
124.1K
James Freitag
James Freitag@jimfreitag·
@aryehazan Ah - my bad - I thought you were talking about like mathematician colleagues. Definitely can believe you heard a lot of wild stuff here on twitter!
English
1
0
0
36
Aryeh Kontorovich
Aryeh Kontorovich@aryehazan·
"it's possible to be a good mathematician with bad politics" this statement will be completely uncontroversial on the Right, but on the Left is often treated as a non-possibility
English
18
3
88
4.3K
James Freitag
James Freitag@jimfreitag·
@Golovanov_ammoc The fundamental theorem of mathematical logic is the completeness theorem, not the incompleteness theorem.
English
2
0
15
350
Yaashaa Golovanov
Yaashaa Golovanov@Golovanov_ammoc·
(Dis)agree for fundamental theorem of ? • Analysis: Spectral Thm • Logic: incompleteness Thm • Combinatorics: PHP • Calculus: Stokes Thm • Topology: rule of product • Geometry: Pythagoras Thm • Probability: Central Limit Thm • Arithmetic: Prime factorisation LMK!
English
18
3
76
15K
James Freitag
James Freitag@jimfreitag·
@littmath @tunguz That seems like such a crazy opinion that I read this as saying that it is completely unrealistic for AI to be writing papers as good as Witten’s in 2 years.
English
0
0
11
1.1K
Daniel Litt
Daniel Litt@littmath·
@tunguz “Ed Witten isn’t brilliant” is certainly a take!
English
8
1
374
17.1K
James Freitag
James Freitag@jimfreitag·
@kareem_carr Motivation is a funny thing. Some people treat math like a sport. Others treat it like an enjoyable hobby, etc. If you enjoy math as a sport, it isn't so surprising that you might enjoy it more if you happen to be quite good at it (which might mean various things).
English
2
1
11
1.4K
Dr Kareem Carr
Dr Kareem Carr@kareem_carr·
I don’t understand giving up on studying math, if you really love it, on the grounds that other people are better. I like gardening, and it would never occur to me to stop because other people grow nicer tomatoes.
English
88
65
1K
51.3K
James Freitag
James Freitag@jimfreitag·
@pmddomingos All but a tiny number of mathematics papers involve proofs that can be placed in one uncontroversial axiomatic system. Most in a very weak fragment of this system. In a practical sense, what you write is not correct.
English
0
0
1
91
Pedro Domingos
Pedro Domingos@pmddomingos·
Mathematics is not science because it’s subjective. Whether a theorem follows from axioms isn’t, but the choice of axioms and which of their consequences get to be called theorems very much is. In contrast, in science reality dictates the choices.
English
224
46
492
186.9K
James Freitag
James Freitag@jimfreitag·
@TobyMeadows @aryehazan ok - yeah I actually know nothing about him and haven’t read his book. I am also very bad at interpreting jokes due mostly to my low iq
English
0
0
1
66
Toby Meadows
Toby Meadows@TobyMeadows·
@jimfreitag @aryehazan I was talking about Aryeh's counterfactual. Still I didn't know this. He seems to be one of those doubly awful people who appear to be so full of wank that they can't possibly be any good and yet they are.
English
1
0
3
95
Aryeh Kontorovich
Aryeh Kontorovich@aryehazan·
I solemnly promise that had I attended the most elite undergrad math institution in the world (the École Normale Supérieure, which produced more Fields medal than the entire continent of Asia), I would not be insufferable about it
English
26
6
247
27.2K
James Freitag
James Freitag@jimfreitag·
@TobyMeadows @aryehazan I mean… the guy published a hundred page annals paper 10 years ago. It’s probably ok for him to claim he knows a bit about math research?
English
2
0
1
86
Toby Meadows
Toby Meadows@TobyMeadows·
@aryehazan Hear me out ... you could also write coffee table books that aren't really about math and still pretend you're a serious researcher. You'd have options.
English
2
0
7
1.9K

Keşfet

@AliceFromQueens @feelsdesperate @lreyzin @JDHamkins @LocasaleLab @joelwatsonfish @rand_longevity @littmath