Trending this week: GPT Neo framework; PixLoc to learn robust camera localization from pixels to pose; More Deep Fake legislation, and advances in Federated Learning.
Every week we analyze the most discussed topics on Twitter by Data Science or AI influencers.
The following topics, URLs, resources, and tweets have been automatically selected by a topic model. Want to know more? Jump at the end of this article!
This Week Overview
This week, Data Science / AI influencers have shared articles about:
- AI Trends For 2021
- State-Of-The-Art Deep Learning Resources
- Atlanta Spa Shootings
Here are all the details for each topic:
AI Trends For 2021
In this section: AI trends projections for 2021 shared by some influencers.
AI has shared an article titled AI and Data Science Predictions For 2021. This article talks about the explosion of AI as-a-Service in businesses, more Deep Fake legislation, and considerable advances in Federated Learning.
On his side, Terence Mills retweeted an article presenting the AI Job Trends Important To Watch in 2021. This article gives a list of jobs that require machine learning skills, and it also describes the evolving skillsets that are in demand in most organizations.
Finally, ipfconline shared a link to a presentation of a list of 100 Artificial Intelligence Startups To Watch in 2021. These startups are qualified as being the world’s fastest-growing AI startups.
State-Of-The-Art Deep Learning Resources
This week, some influencers have shared new resources on state-of-the-art Natural Language Processing (NLP) and Computer Vision (CV).
Aran Komatsuzaki has shared a link to a GPT Neo Github repository, which provides “an implementation of GPT-2 & GPT-3-like models with the ability to scale-up full GPT-3 sizes, using the mesh-TensorFlow library”.
He also shared a research paper presenting P-tuning, “a novel fine-tuning method that allows GPTs to achieve comparable and sometimes better performance to similar-sized BERTs in supervised learning.”
Thomasz Malisiewicz has shared a research paper presenting PixLoc, a scene-agnostic neural network that learns robust camera localization from pixels to pose. PixLoc “learns strong data prior by end-to-end training from pixels to pose and exhibits exceptional generalization to new scenes by separating model parameters and scene geometry.”
Atlanta Spa Shootings
On March 16, 2021, a series of mass shootings occurred at three spas or massage parlors in the metropolitan area of Atlanta, United States. This stirred a response from the data science community on Twitter.
Ayodele retweeted inciting followers to go beyond tweeting and dismantling the anti-Asian mindset. She also retweeted to invite followers to fight for all groups who are targeted by white supremacy at the same time — be it Black or Asian.
In order to find Twitter most discussed topics within the data science / AI community, we created a whole pipeline combining influencers analysis, data extraction, and NLP using the BERTopic Python library, a topic modeling technique that leverages:
- Sentence Transformers, to obtain a robust semantic representation of the texts
- HDBSCAN, to create dense and relevant clusters
- Class-based TF-IDF (c-TF-IDF), to allow easy interpretable topics whilst keeping important words in the topics descriptions.