Love this paper. "At what cost is it worth doing simulation?" is a really critical one in an age of sampling thousands of compositions in a day with self-driving and high-throughput labs.
Great new work by Ryan Jacobs, Philip E Goins and Dane Morgan @UWMadisonMSE @UWMadison @ArmyResearchLab - 'Role of multifidelity data in sequential active learning #materialsdiscovery campaigns....' - iopscience.iop.org/article/1… #machinelearning #materials #AI #compchem #condmat
Joseph Montoya retweeted
New Scientists covered some of our recent work today.
"There is a line of sight to the time when we will have research assistants that are AIs that have access to incredible amounts of information"
US government lab is using GPT-3 to analyse research papers newscientist.com/article/235…
Some really incredible results using new LLMs from @jmdagdelen, @jainpapers, @ardunn_pro, and team. Data extraction from the literature is seeming more and more like a solved problem.
Really excited to share some of our new work on scientific NLP:
We demonstrate that LLMs like #GPT3 can easily be trained to complete challenging scientific information extraction tasks with just a few hundred examples.
arxiv.org/abs/2212.05238
Demo: matscholar.com/info-extracti…
11
Really enjoyed chatting with @taylordsparks and Andrew Falkowski on the Materialism Podcast. We had a great conversation about catalysts, the ways they've shaped society, and what the future might hold. Check it out!
From food production to #EVs, #catalysts have spurred many global developments.
TRI's Joey Montoya shares how Toyota is using catalysts to address climate change on the latest episode of @MaterialismPod with @taylordsparks.👂👇
materialismpodcast.com/episo…
#BEVs #FuelCells
Check out Ha-Kyung Kwon's highlight of @ToyotaResearch's release of an open-source polymer database focused on battery electrolyte properties. Overcoming the challenges of collecting, organizing, and distributing polymer data is a big achievement! medium.com/toyotaresearch/bu…
Had the privilege of speaking at the #cat70norskov symposium a couple of days ago. Like everything that Jens does, his approach to advising is beautifully simple - when I presented results he would always ask me two questions: what does a result say, and then what does it mean?
The best thing about this symposium was seeing not only the results of Jens' pedagogy on former students, but on second generations being trained by his first. The challenges we face in catalysis are critical for solving climate change, but this gives me hope for the future.
1
2
It's also a big reminder of how impactful education and mentorship can be. For more on Jens' career, check out this great perspective that @medford_group and @AndyAPeterson wrote at pubs.acs.org/doi/abs/10.1021…
1
Joseph Montoya retweeted
Introducing the world’s first nanomaterial "data factory” – a novel AI-driven technique by @ToyotaResearch & @NorthwesternU to uncover new materials faster to meet the growing demand for low-carbon mobility.
See full video and learn more:
pressroom.toyota.com/?p=7777…
#TRIxNU
Joseph Montoya retweeted
Introducing Drift Dream! 👀🏎️⤵️
Designed by TRI to autonomously drift around obstacles, the @Toyota GR Supra is now available as a @Hot_Wheels collectible.
Learn more: youtube.com/watch?v=otO4i_9f…
#AutonomousDriving #hotwheels
The inimitable Handong Ling wrote a very nice XRD-solving framework while an intern at TRI, which was published today! You can get the code at github.com/TRI-AMDD/CAMD-XRD and read the paper at
doi.org/10.1016/j.commatsci.….
Preprint available at: chemrxiv.org/engage/chemrxiv…
1
An overview of TRI-supported fuel cell catalyst research was just published in @ees_journal. If you're interested in perspectives on organizing a research consortium, on why future ORR science should focus on durability, or how AI fits in, check it out!
doi.org/10.1039/D2EE01333A
Joseph Montoya retweeted
We’re very excited to release the Open Catalyst 2022 (OC22) Dataset: Oxide Electrocatalysis!
Paper: arxiv.org/abs/2206.08917
Dataset: github.com/Open-Catalyst-Pro…
Blog: ai.facebook.com/blog/acceler…
1/n
First batch of results from our autonomous DFT system, CAMD, released earlier this week, ~100k diverse structures, ~25k metastable. Excited about trying to make some of the new structures, but the "bad" data may be even more useful for ML/AL.
Read it at:
nature.com/articles/s41597-0…
Joseph Montoya retweeted
Them: Ownership is purposely trying to make the experience at the Coliseum terrible so fans won’t come.
Me: THEN EXPLAIN THIS.
#DrumTogether
Joseph Montoya retweeted
In the past decade, many new tools emerged to help us better design new materials - it's my pleasure to write about our TRI team's recent work and my own experience moving towards autonomous research.
By automating scientific processes and introducing #artificialintelligence for decision-making, TRI’s new #closedloop research platforms free up scientists’ time for more creative tasks.
TRI's @SunShijing demonstrates in her latest Medium blog post. ⤵️
medium.com/toyotaresearch/ho…












