Wes Kendrick

Machine Learning Engineer @AWS. 🇺🇸 in 🇪🇺. Building (and writing about) tools, systems, and interfaces of all kinds.

[Cédric Archambeau 🇺🇦🇪🇺🦋 @cedapprox]

Congratulations @kleiaaro , @j_golebiowski , and @weskendrick ! Very proud of the impact of the Berlin #AWS AI science teams today!

[Wes Kendrick @weskendrick]

@innixma Congrats on the release!

Love the detailed release notes 💪

[Wes Kendrick @weskendrick]

Excited to see what folks build with https://t.co/x8rX3Ymfjz!

[Jonathan Weiss @jweiss]

PartyRock is live! Go checkout this fun & easy way to try out various GenAI models, learn about LLMs, and experiment by building small apps:

https://t.co/wJcDYAKuok

Amazing work @TLobinger & team!

[Wes Kendrick @weskendrick]

Thank you to everyone who contributed to the release, including David Salinas (@nonameyet8), Matthias Seeger, Aaron Klein (@kleiaaro), and Jacek Golebiowski (@j_golebiowski)

[Wes Kendrick @weskendrick]

We just released Syne Tune v0.10.0!

✅Integration with SMAC optimizer
✅New example: Tune XGboost
✅New example: Resume Tuning

Release notes: https://t.co/OOKehYg5jC

[Wes Kendrick @weskendrick]

Impressive demonstration of how far technology has advanced

https://t.co/l70JvvKjSY

[Wes Kendrick @weskendrick]

If a team manages to do this, it would be an impressive demonstration of how far technology has come

[Wes Kendrick @weskendrick]

Amazing work 👏👏👏

[Nat Friedman @natfriedman]

Today we are announcing a major breakthrough in the Vesuvius Challenge: we have read the first word from an unopened Herculaneum scroll.
The word is "πορφυρας" which means "purple dye" or "cloths of purple."

https://t.co/mSbHtzNbAl

Congratulations to 21yo computer science student @LukeFarritor who is the first person to see this handwriting in nearly 2000 years. He has won the $40,000 First Letters prize for this world-historical achievement.

We are also awarding a $10,000 First Ink prize to @CJHandmer who was the first person to see ink and multiple letters within an unopened scroll. His work was the basis of Luke's ML model.

And @Youssef_M_Nader has won a $10,000 second-place First Letters prize for producing the clearest and most comprehensive images from inside a scroll yet.

This has been the dream of many people since the scrolls were first discovered in the 1750s. It is also the result of 20 years of work from Dr. Brent Seales and his team at EduceLab, whose years of dedicated work have made this last mile possible.

The $700,000 Vesuvius Challenge Grand Prize is now in sight. Who will claim it?

[Joaquin Vanschoren @joavanschoren]

Two more great AutoML keynotes at @automl_conf: Awesome practical advice by @eytan Bakshy on efficient optimization of living ML systems, and the ever-inspirational Esteban Real on evolving AI models from scratch vs. giving prior knowledge. Here are my quick notes :) #AutoML2023

[Wes Kendrick @weskendrick]

@nabeelqu Paul-Ernst-Park (~1hr outside Berlin via S-Bahn, nice trail around the lake + Biergarten)

Enjoy your visit!

[Wes Kendrick @weskendrick]

Anyone maintaining a similar list for Berlin/Europe?

[Charlene Li (she/her) @charleneli]

Want to attend more AI-related events in your area? Check out Cerebral Valley. It’s my ultimate AI resource for events in SF or New York. Some weeks have 20 different meetups! (Warning: it may induce serious FOMO.) https://t.co/VhpBQXxzhZ

[Wes Kendrick @weskendrick]

Who knew multiplication could be so interesting!

arXiv link to their paper, check it out: https://t.co/37sUH0XCAK

[Wes Kendrick @weskendrick]

The authors found that off-the-shelf ChatGPT/GPT4 achieves less than 60% accuracy on 1-digit by 1-digit to 5-digit by 5-digit multiplication problems such as "12345 times 12345"

Just for fun, I prompted ChatGPT with a handful of multiplication problems. Here's what I saw: 🧵

[Nouha Dziri @nouhadziri]

🚀📢 GPT models have blown our minds with their astonishing capabilities. But, do they truly acquire the ability to perform reasoning tasks that humans find easy to execute? NO⛔️

We investigate the limits of Transformers *empirically* and *theoretically* on compositional tasks🔥

[Wes Kendrick @weskendrick]

More broadly:
1. Will incremental improvements in LLMs "solve" multiplication? (e.g. bigger models, better data), or will new architectures be needed?
2. Can today's transformers reliably use external tools such as calculators? Under which situations does tool use break?

5/

[Wes Kendrick @weskendrick]

Unit tests also run about 30% faster now 🎉

[Jacek Golebiowski @j_golebiowski]

New Syne Tune release! Top new features:

1. BO with SKLearn estimators as surrogates
2. Conformal Quantile Regression HPO from ICML2023
3. New Multi Surrogate Multi Objective Searcher
4. Code for OTHPO by Hellan et al.
5. New tutorials

https://t.co/GbRFESMXnZ details below 🧵

[AutoML Seminar @AutomlSeminar]

Join us on Thursday 4pm CEST with @The_Eimer to talk about "Challenges in Hyperparameter Optimization for Reinforcement Learning - Understanding AutoRL Through an AutoML Lens'': https://t.co/jE5xCOPfof

[Wes Kendrick @weskendrick]

This team ships!

[Lukas Balles @lukas_balles]

We just released v0.3.0 of Renate, our PyTorch library for continual learning. We now have improved support for NLP applications and provide methods for distribution shift detection, which can help you decide when to update your model.

https://t.co/Z6g3WhQyTP

[Wes Kendrick @weskendrick]

@altryne @AviSchiffmann @swyx @KolinKoehl Interesting to hear it summarize the comments section.

Watch out for people trying to add prompt injections in the comments though now 😬