David Mimno

2.4K posts

David Mimno

David Mimno

@dmimno

he teaches data science at Cornell

Katılım Haziran 2009
2.1K Takip Edilen4.8K Takipçiler
Sabitlenmiş Tweet
David Mimno
David Mimno@dmimno·
Optimist: AI has achieved human-level performance! Realist: “AI” is a collection of brittle hacks that, under very specific circumstances, mimic the surface appearance of intelligence. Pessimist: AI has achieved human-level performance.
English
75
3.5K
9.3K
0
David Mimno retweetledi
Michael Saxon
Michael Saxon@m2saxon·
So, your model beats a random baseline (1/n class) in dev, so you know it's good right? WRONG @gyauney and @dmimno : since you gave your model/ICL prompt/exemplars many tries on dev, you need to give *the random baseline* many tries too! 1/
Michael Saxon tweet media
English
1
5
21
3.1K
David Mimno retweetledi
Sasha Rush
Sasha Rush@srush_nlp·
If for whatever reason you can't make it to COLM, please send me an email, and we will give you a refund. There are a lot of people who would really like to come.
English
1
7
59
16.1K
David Mimno retweetledi
Jenn Wortman Vaughan
Jenn Wortman Vaughan@jennwvaughan·
Fantastic opportunity for undergrads interested in research! I've been lucky to host and collaborate with several amazing interns through this program. Deadline 10/21. Help spread the word! 📢📢📢
Microsoft Research@MSFTResearch

Our Undergraduate Research Internship Program is now accepting applications for summer research internships. Rising college juniors & seniors who are passionate about technology and champion diversity and inclusion are encouraged to apply by 10/21: msft.it/6017meOVF

English
0
10
18
4.9K
Zachary Lipton
Zachary Lipton@zacharylipton·
Pattern worth prying open: 1. Set of publishing norms makes a field open & dynamic, attracts the weird dreamers. 2. Field produces technology, gets hot, hordes of brain-dead opportunists descend. 3. Same open systems that made field dynamic now buckle under load, bad faith.
Zachary Lipton@zacharylipton

Hard to say where we go from here, but papers in our now-bloated conferences have devolved into such gibberish, and reviews have become so haphazard that publishing imprimaturs in ML are truly meaningless.

English
4
2
41
8.5K
David Mimno retweetledi
Mark Dredze
Mark Dredze@mdredze·
The early 🦜 gets the 🪱 @JHUCompSci has a great opportunity for faculty hiring. Apply early and you couls interview early and 🤞get an early offer. Spreading out interviews and having an early offer = less stressful job search.
Mark Dredze@mdredze

🚨 Johns Hopkins @JHUCompSci is hiring faculty at all ranks! 1) Data Science and AI 🤖; 2) All other areas of CS 💻. We will double in size to become one of the largest departments in the coming years! 🚀 Note: early action and spousal placement. 👥 🧵 apply.interfolio.com/153420

English
2
18
61
15.9K
David Mimno retweetledi
Karen Levy
Karen Levy@karen_ec_levy·
Returning from perpetual Twitter hiatus to spread the word: @CornellInfoSci is hiring! Tenure-track hires at all levels in Ithaca, focusing on qual/interpretive methods, CSS/quant approaches, study/design of human/AI interactions. Review starts 11/10: academicjobsonline.org/ajo/jobs/28405
English
2
51
119
29K
David Mimno
David Mimno@dmimno·
I have a very particular set of skills. Skills that I have developed over a long career. Skills that make me a nightmare for 11-page pdfs like you. If you go down to 8 pages now, that will be the end of it.
English
3
0
27
2.2K
David Mimno
David Mimno@dmimno·
Open “source” is a relic of the 90s era separation between code and executable. Retrofitting it is hopeless. The Creative Commons license family is a better model: describe aspects like open weights, protocol, data with succinct codes.
Percy Liang@percyliang

Thoughts on OSI’s *draft* open-source AI definition v0.0.9: Key question: how much information about the data needs to be disclosed to get the open-source stamp? My initial reaction was that the entire dataset must be released, but now I think this is neither sufficient nor necessary.

English
1
1
15
1.9K
David Mimno
David Mimno@dmimno·
@thechrisperry Colab is so valuable as a teaching tool! Especially for students who can’t get anaconda to work or don’t have disk space
English
0
0
1
128
David Mimno
David Mimno@dmimno·
@thechrisperry A system that embeds a different model trained to emulate a TA could be an option: provide hints and point out bugs but not runnable code. It would have to be opt-in for students who want to learn, there’s no way to prevent people who want to get answers.
English
2
0
2
186
David Mimno
David Mimno@dmimno·
PSA: if you assign python notebooks for homework, you CANNOT allow students to use Google colab anymore, it reads instructions and autosuggests good answers. I’m trying “You’ve earned the right not to play on easy mode”, but the temptation to just say “that looks good” is there
English
4
8
52
10.8K
David Mimno
David Mimno@dmimno·
@arthur_spirling Moving towards that, sure, but I want them to have a base familiarity that allows them to evaluate what they’re being given
English
0
0
6
711
Arthur Spirling
Arthur Spirling@arthur_spirling·
@dmimno I hear you, but isn’t this just a strong signal that the pedagogical value in this sort of stuff without autocomplete is p minimal now? I mean, in terms of prepping them for “real life” etc
English
1
0
3
829