David Mimno
@dmimno
Followers
5K
Following
7K
Media
73
Statuses
2K
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.
77
4K
9K
This is big. Nvidia is probably ahead a generation on chips, but CUDA is the moat.
Now you can use bitsandbytes on AMD GPUs and Intel hardware. This is a big milestone and was a huge undertaking. @Titus_vK did an amazing job here. Eager to hear feedback! Let us know how it works for you.
0
0
6
RT @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….
0
7
0
RT @jennwvaughan: Fantastic opportunity for undergrads interested in research! I've been lucky to host and collaborate with several amazing….
0
10
0
More evidence Tech companies’ spending on GPUs and electricity is crowding out hiring. Hardware vs HR. H100s will become e-waste, interns will become CEOs.
Last week, the WSJ reported on the decline of tech jobs, particularly entry-level and recruiting roles, suggesting they might be gone for good. I initially thought it was exaggerated, but then saw a UC Berkeley CS professor mention that even his top students are struggling to get.
0
2
7
RT @mdredze: The early 🦜 gets the 🪱. @JHUCompSci has a great opportunity for faculty hiring. Apply early and you couls interview early and….
0
19
0
RT @mellymeldubs: We're hiring at @UW_iSchool! 🦭🗻🏙️ Come join us in Seattle. 2 x tenure-track Assistant Prof in Informatics: .
0
31
0
RT @nolauren: Debates in #DH Computational Humanities is now available for purchase (OA @ManifoldScholar this winter)! Thank you to the ama….
0
89
0
RT @msalganik: Department of Information Science at Cornell; open rank faculty search. Field is open and there is a special emphasis in a f….
0
31
0
RT @whynotyet: We are hiring at @CornellInfoSci ! Open rank: Cornell is a fantastic place to work, please consider….
0
31
0
RT @karen_ec_levy: Returning from perpetual Twitter hiatus to spread the word: @CornellInfoSci is hiring! Tenure-track hires at all levels….
0
51
0
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.
3
0
28
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.
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.
1
1
15
RT @yian_yin: Truly excited to be featured in this @NorthwesternEng article! Shoutout to many at @NU_IEMS and @NICOatNU who've been part of….
0
2
0
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.
4
8
53
What I love about these is that I have spent so much time reading equations in papers and textbooks just trying to answer “what does it actually do”, and these just make it unmistakably obvious.
[Backpropagation] by Hand✍️. [1] Forward Pass.↳ Given a multi layer perceptron (3 levels), an input vector X, predictions Y^{Pred} = [0.5, 0.5, 0], and ground truth label Y^{Target} = [0, 1, 0]. [2] Backpropagation.↳ Insert cells to hold our calculations. [3] Layer 3 -
0
0
17
RT @manoelribeiro: This is huge: Reddit4researchers just opened a beta program! If you are interested about Reddit, apply to gain cool API….
0
43
0
Can we all remind ourselves that the point of this is to learn about neural network training, and that using a billion flops to badly simulate what is literally one (1) flop is an insane idea? I feel like this gets lost.
learned something super interesting this week. if you train GPT-2 to multiply, you can't even train it to multiply 4-digit numbers (30% accuracy). but if you use a really clever (and somewhat complex) training scheme, GPT-2 can generalize up to 20-digit numbers (100% accuracy).
2
0
26