#ComplexSystems folks:
After more than a decade, I am planning to update this diagram. What kind of keywords, concepts, topics, and/or research areas should be added? Let me know
commons.wikimedia.org/wiki/File:Comp
Joshua Nemecek’s Tweets
Are all human languages equally complex? 👇
Meta-analyses of 28 complexity metrics applied to texts written in overall 80 typologically diverse languages:
- partially support the trade-off hypothesis
- fully support the equi-complexity hypothesis
degruyter.com/document/doi/1
24
250
940
Show this thread
Our attempt at a #betterposter at #emnlp2022. We had great engagement. Some people were content to scan a code from the back, while the detailed portions were useful in discussions and questions. It was also easier to make.
When Jim Cantore is calling it horrendous, you think he's coming to Michigan?
Quote Tweet
It really will take till Christmas morning for things to settle out. Otherwise Thursday through Saturday looks horrendous over the Great Lakes, Northeast and Mid-South with rapidly changing and dangerous weather elements.
GIF
While I didn't come across it in my research prior to the conference, this is insightful and valuable as well: spectrum.library.concordia.ca/id/eprint/9865
Show this thread
RAIL (Responsible AI Licenses) licenses.ai offer a way to prevent harmful use (including violations of Data Sovereignty) (ours is here: huggingface.co/spaces/sil-ai/)
1
Show this thread
Cristina Voinea's "Designing for conviviality" parallels with concepts of Data Sovereignty isis.wien/wp-content/upl
1
Show this thread
Also Gary Simons, et al. on "Assessing Digital Language Support on a Global Scale" arxiv.org/abs/2209.13515
1
Show this thread
My colleagues M. Paul Lewis and Gary F. Simons on "Sustaining Language Use"
1
Show this thread
In the area of Language Ecology, I learned from on "Decolonizing Speech and Language Technology" aclanthology.org/2020.coling-ma
1
1
Show this thread
Papa Reo papareo.nz is a strong example of Data Sovereignty, covering Te Reo Māori and Pacific Island languages
1
Show this thread
1
Show this thread
Also looked at "Decolonial AI" arxiv.org/pdf/2007.04068
2
Show this thread
builds on open source "FAIR" principles with "CARE" for Indigenous Data Governance
1
Show this thread
provides training on Data Sovereignty specifically for First Nations
1
3
Show this thread
data labels help to describe content with metadata to prevent misuse localcontexts.org/labels/traditi
1
Show this thread
I'm giving a presentation with the title "Towards ecologically-aware language technology that promotes sustained language use" at the conference on Saturday, and wanted to pre-Tweet out some of the references I used.
3
4
7
Show this thread
I don’t train from scratch, I use RoBERTa🧐
Wait…
Why not cross-encoder/stsb-roberta?facebook/muppet-roberta?
We automatically identify the best models on 🤗(periodically)
Just pick the best one
and finetune on your task
ibm.github.io/model-recyclin
8
66
261
Show this thread
Is data really important for pretraining?
Could we just pretrain on 1 picture? Only synthetic text? Fractals?
A 🧵 summing the image and text papers that do just that.
and they all have a similar conclusion🤔
7
72
357
Show this thread
Today we’re excited to announce our partnership with , , , and to create the first open source reinforcement-learning based instruction tuned language model. You can read more about our efforts here: carper.ai/instruct-gpt-a 1/4
8
144
587
Show this thread
The denoising objective which created state of the art image generation models is being adapted to produce SoTA Large Language Models. Cool. Now do audio LLMs.
Quote Tweet
Introducing UL2, a novel language pre-training paradigm that improves performance of language models across datasets and setups by using a mixture of training objectives, each with different configurations. Read more and grab model checkpoints at goo.gle/3euHrEo
GIF
read image description
ALT
A implementation of text-to-3D dreamfusion, powered by stable diffusion
github: github.com/ashawkey/stabl
0:25
138.3K views
27
563
1,995
Show this thread
The "What do you want to be..." question typically encourages goals over values. Goals imagine future possibilities based on our present expectations. Values prepare us for a changing and uncertain world.
Quote Tweet
True innovation must be inclusive.
#SpeechRecognition is no exception. We’re working with to enable language and accent recognition at the #edge, creating a future solution for everyone.
2
11
21
Delighted to announce the public open source release of #StableDiffusion!
Please see our release post and retweet! stability.ai/blog/stable-di
Proud of everyone involved in releasing this tech that is the first of a series of models to activate the creative potential of humanity
147
2,119
5,638
Show this thread
Finally! We can stop talking about Tay. This is progress. Lots of cringe examples in this thread. twitter.com/JeffHorwitz/st
This Tweet is unavailable.
This is a milestone for multilingual large language models.
Quote Tweet
▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓ 100%
It turns out Lytro was ahead of its time. Take a photo first, then focus. Although Lytro was bought by Google who created this, so in a way, it continues.
Quote Tweet
If you're still at CVPR and have the stamina to make it through another poster session, check out RawNeRF tomorrow morning! We exploit the fact that NeRF is surprisingly robust to image noise to reconstruct scenes directly from raw HDR sensor data.
Show this thread
1:00
67.9K views
5
122
591
I started the game Titanfall 2 tinyurl.com/jl7ttcz
1
I started the game The Elder Scrolls V: Skyrim Special Edition tinyurl.com/hzkg85j
1
Free trade can be wonderful if you have the power of nuclear weapons.
2
161
240
Also, what is wrong with much of Christian media. amzn.com/k/escueSH0RDSK Twelve centuries later, French thinker Blaise Pascal asked, ...
Nate Silver messed up on this table. Look at the lower right of the top table. amzn.com/k/m_RfVZZMQ5-k postseason.