Joined November 2014
- Tweets602
- Following507
- Followers497
- Likes4.7K
Kostas Stathoulopoulos retweeted
💫 Perspective-taking prompting
A new example by @kstathou, which uses SimToM and structured generation.
outlines-dev.github.io/outli…
Kostas Stathoulopoulos retweeted
Try @climatepolradar's new tool:
🌍Search full text of all climate laws & policies from all 200 govts
❓Enter Search Term
🌵Get results for your search + semantically related terms
🧭Navigate to passages that contain your search term.
Here, now, free:
app.climatepolicyradar.org/
Kostas Stathoulopoulos retweeted
We are hiring — come and join the @deepsciventures Climate team! you should have a deep technical understanding of chemistry, ideally also chemical engineering and materials science and the drive to see the impact of your work through venture creation. apply.workable.com/deep-scie… /1
Kostas Stathoulopoulos retweeted
Our latest Fellow Spotlight is now out 🚨
Join us as we discuss the Fellowship experience, the importance of openness in climate action, and @irinimalliaraki's experience working at the intersection of climate and data.
👉 subak.org/blog/fellow-spotli…
Kostas Stathoulopoulos retweeted
I've been working with @protontypes and supported by @ClimateSubak to understand the impact and potential of open source software on climate technology🌳. Below are some early findings and ways you can get involved:
opensustain.tech/blog/impact…
Super excited about OpenAlex. The documentation looks great and splitting the data into 5 scholarly entities eases the ingestion of subsets. It also comes with a permissive data license which is neat.
OpenAlex is live!
Docs are up, API is running, snapshots are available -- dig in and let us know what you think!
Sampler:
Docs: docs.openalex.org/
API: api.openalex.org/works/W2741…
Display: openalex.org/W2741809807
Feeling lucky? explore.openalex.org/
Thread for highlights!
We need metadata to group/filter those interfaces by function, domain, visual elements and learning curve!
colinmegill.com/explorers/
I am going to try to finish one last 2021 task before new years in preparation for teaching interactive data viz @UW
sorting, documenting, and publishing the (nearly, or over) 1000 examples of data exploration I have collected over the past decade
1
1
Kostas Stathoulopoulos retweeted
From 37,380 volunteers, our largest ever crowdsourced study into YouTube’s algorithm shows that @YouTube *STILL* sometimes recommends videos that violate its own policies
Learn more about #YouTubeRegrets in a News Byte edition of this week’s #MozNewsBeat foundation.mozilla.org/en/bl…
Kostas Stathoulopoulos retweeted
📺How to train state-of-the-art sentence embeddings? 📺
Just uploaded my 3-part video series on the theory how to train state-of-the-art sentence embedding models:
youtu.be/qmN1fJ7Fdmo
Between this and Public Health England using excel as a database, the public sector doesn't need more data scientists and "AI" but data engineers and integration tests.
This quoted post is unavailable.
4
Kostas Stathoulopoulos retweeted
Thanks to @MSFTResearch for providing the Microsoft Academic Graph (MAG). We’ve been working with MSR and MAG since 2018, and we’ve been collaborating on this transition for some time. 1/3
microsoft.com/en-us/research…
Kostas Stathoulopoulos retweeted
We revisited our popular 2019 analysis with new data and found that China will soon surpass the US in publishing the top 10% most-cited AI papers, and become the leader in the top 1% of papers in 2023.
Learn more in this article by @jeo_yang & @etzioni:
medium.com/ai2-blog/china-is…
Kostas Stathoulopoulos retweeted
How do search engines match queries to documents, people, or products?
I studied some examples & found three methods:
• Lexical: Rewriting, expansion, relaxation
• Graph: Mapping to concepts & relationships
• Embedding: Via (self-)supervised learning
eugeneyan.com/writing/search…
Kostas Stathoulopoulos retweeted
Today we celebrate the anniversary of Italy's liberation from fascism. So for this day I created a map with almost all Italian streets named after people who resisted and fought fascism, & their Wikipedia pages. Le strade della resistenza. Buon 25 Aprile❤️
mirrys.github.io/mappe_parti…
Kostas Stathoulopoulos retweeted
All 1.8M papers in arXiv - title + abstract embedded with SPECTER (@ai2_allennlp) -> UMAP (@RAPIDSai). Colored by the year of the "update" date, which I think roughly maps to the original upload date.
Kostas Stathoulopoulos retweeted
Technology is shifting to voice-enabled products, but today’s devices cannot understand vast swaths of the world’s languages.
#CommonVoice 🗣️ aims to level the playing field, and our new partnership with @NVIDIAAI will see it grow and expand.
Read more👇
mzl.la/3dckTF3
Kostas Stathoulopoulos retweeted
Open Syllabus is a wonderful project that already collected 9 million English-language syllabi from 140 countries. You can search by title and author, and refine by schools, fields and countries. Check our their amazing interactive data visualization -zcu.io/Btu3
Kostas Stathoulopoulos retweeted
Last year, we did a rapid review of the evidence for good group decisions to help us work more effectively @nesta_uk. Here's what we found, in case it's useful for others who want to make the most of their teams' collective intelligence! tinyurl.com/78je5zyu
Kostas Stathoulopoulos retweeted
"Deep Learning, Deep Change", our analysis of the geography of deep learning research and the drivers of DL research clustering, is finally out in Scientometrics.
link.springer.com/article/10…
