AI models work best for English, and well for a handful of other languages—meaning the technology may leave behind thousands of tongues and billions of people who speak them. Some languages could go extinct.

I explore why, and global efforts to change that, for The Atlantic:

AI understands just a tiny fraction of the thousands of languages used worldwide. That means the promised chatbot revolution could shut out billions of people, and even push some languages to extinction, Matteo Wong writes: theatlantic.com/technology/arc…

Have a read at my recent intervention on the Atlantic media !!!

theatlantic.com/technology/arc…

🧵1/6 We're so excited to announce the inaugural machine learning for drug discovery summer school!

🏫 What: Lectures from Yoshua Bengio, Anne Carpenter, PhD, Pratyush Tiwary, Bharath Ramsundar, and more.

📍Where: MilaQuebec

📅 When: June 12th - 21st

Learn more here: portal.ml4dd.com

I'll be attending ICLR 2024. I'll be co-hosting the #AfricaNLP workshop with folks from Masakhane & I'll be a panelist at the 'Practical ML for Low-Resource Settings' workshop, discussing resource efficient development and use of LLMs. Looking forward to great discussions.

We're excited to launch our newsletter!🎊Your monthly dose of our latest product news and our moves.

The first edition of our newsletter covers our product Vulavula, its features, the latest product improvements and more.

Read and Subscribe here: bit.ly/LelNews

Does anyone know a book cover designer ? Please reach out 🙏🏿

🥲Sad, very sad

Full house for my lecture hosted by the Center of Mathematical Sciences and Applications at Harvard yesterday.

Video will be available from the CMSA YouTube channel.
Slides here: drive.google.com/file/d/1Ymx_LC…

#RayBanMeta

We present you the world’s Best Twi Text to Speech System. Made by Ghanaians for Ghanaians. Try it today.

Hit my first #SIGUL paper with almost a perfect review 😅🔥 Unfortunately I won’t make it to LREC COLING 2024 but I’m excited to share virtually my work. Stay tuned for more updates

Have been recently studying more about world model and realize this 'old' paper by Prof. Jürgen Schmidhuber
resonated a lot with recent trend such as model based SSL, JEPA and digital twin etc.. Will share it with students in the class here in Asia
arxiv.org/abs/1803.10122

I do not believe human-level AI (artificial superintelligence, or the commonest sense of #AGI ) is close at hand. AI has made breakthroughs, but the claim of AGI by 2030 is as laughable as claims of AGI by 1980 are in retrospect. Look how similar the rhetoric was in LIFE in 1970!

A recent work by Quentin Garrido and Yann LeCun used an world model ( transition model) to enhance self-supervised learning. It is quite impressive. Just one confusion, the action/intervention used are not 'real world', any way to improve on that front? arxiv.org/abs/2403.00504

EMNLP 2024 will take place in Miami, Florida from Nov 12th to Nov 16th, 2024, at the Hyatt Regency Miami Hotel. More information: 2024.emnlp.org #EMNLP2024

Tanmoy Mukherjee Cfp is out now. No abstract deadline, just ARR June. 2024.emnlp.org/calls/main_con…

I made some new class slides on “a tour of modern LMs” that has some observations about characteristics of recent LLMs, mostly focusing on open LLMs where we know their details: phontron.com/class/anlp2024…

Check it out if interested, and feedback is welcome!

It's not how big your data is, it's what you do with it.

cc @[email protected] on Mastodon Joana Varon

pml4dc.github.io/iclr2023/pdf/P…