Week Ending 1.9.2022
RESEARCH WATCH: 1.9.2022
This week was very active for "Computer Science - Artificial Intelligence", with 175 new papers.
The paper discussed most in the news over the past week was by a team at DeepMind: "Ethical and social risks of harm from Language Models" by Laura Weidinger et al (Dec 2021), which was referenced 7 times, including in the article How bias creeps into large language models in Analytics India Magazine. Melanie Mitchell (Santa Fe Institute), who is not part of the study, said "[The] ways that we measure performance of these systems needs to be expanded … When the benchmarks are changed a little bit, they [often] don’t generalize well". The paper got social media traction with 61 shares. The researchers aim to help structure the risk landscape associated with large - scale Language Models (LMs). On Twitter, @TobyWalsh said "The first word is misplaced. The paper is not comprehensive. 30 pages of risks. 3 pages on possible means to mitigates those risks. If it were comprehensive, it would cover mitigation in as much detail as harms".
The paper shared the most on social media this week is by a team at The Hong Kong University of Science and Technology: "Speech-to-SQL: Towards Speech-driven SQL Query Generation From Natural Language Question" by Yuanfeng Song et al (Jan 2022) with 94 shares. The investigators work towards designing more effective speech - based interfaces to query the structured data in relational databases. @omarsar0 (elvis) tweeted "Speech-to-SQL - a neural network to translate human speech to SQL statements. If very precise, this could really improve the productivity of data scientists. I wonder how Copilot does on this task. I prefer Text-to-SQL but this is interesting".
Over the past week, 200 new papers were published in "Computer Science - Computer Vision and Pattern Recognition".
The paper discussed most in the news over the past week was "Masked Autoencoders Are Scalable Vision Learners" by Kaiming He et al (Nov 2021), which was referenced 7 times, including in the article Recent advances in dealing with data size challenges in Deep Learning in Towards Data Science. The paper got social media traction with 990 shares. On Twitter, @giffmana commented "1/N The return of patch-based self-supervision! It never worked well and you had to bend over backwards with ResNets (I tried). Now with ViT, very simple patch-based self-supervised pre-training rocks! First BeIT, now Masked AutoEncoders i1k=87.8% 🧶".
Leading researcher Luc Van Gool (Computer Vision Laboratory) came out with "Sound and Visual Representation Learning with Multiple Pretraining Tasks" @ak92501 tweeted "Sound and Visual Representation Learning with Multiple Pretraining Tasks abs: Multi-SSL surpasses recent methods such as MoCov2, DenseCL and DetCo by 2.06%, 3.27% and 1.19% on VOC07 classification and +2.83, +1.56 and +1.61 AP on COCO detection".
The paper shared the most on social media this week is by a team at Tsinghua University: "Vision Transformer with Deformable Attention" by Zhuofan Xia et al (Jan 2022) with 84 shares. @ducha_aiki (Dmytro Mishkin) tweeted "Vision Transformer with Deformable Attention Zhuofan Xia, Xuran Pan, Shiji Song, Li Erran Li, Gao Huang tl;dr: use billinearly interpolated features in key and values".
Over the past week, 24 new papers were published in "Computer Science - Computers and Society".
The paper discussed most in the news over the past week was "Treehouse: A Case For Carbon-Aware Datacenter Software" by Thomas Anderson et al (Jan 2022), which was referenced 2 times, including in the article [ January 7, 2022 ] Wanted: An Energy-Aware Datacenter Application Scheduler Compute in The Next Platform. The paper got social media traction with 7 shares. A Twitter user, @schemeprincess, posted "New research vision paper just dropped on arXiv! Super excited about this new research direction on carbon-aware datacenter software with Tom Anderson, Adam Belay, Mosharah Chowdhury and funded by the and".
This week was active for "Computer Science - Human-Computer Interaction", with 25 new papers.
The paper discussed most in the news over the past week was "Towards a Science of Human-AI Decision Making: A Survey of Empirical Studies" by Vivian Lai et al (Dec 2021), which was referenced 1 time, including in the article Best of arXiv.org for AI, Machine Learning, and Deep Learning – December 2021 in InsideBIGDATA. The paper got social media traction with 53 shares. On Twitter, @QVeraLiao said "Sharing our survey paper, summarizing and reflecting on the study design choices in 100+ papers of empirical studies on human-AI decision making. One of the most fun projects I did in 2021 😇 made an awesome summary thread 👇(and she is on the job market!)".
This week was active for "Computer Science - Learning", with 251 new papers.
The paper discussed most in the news over the past week was by a team at DeepMind: "Improving language models by retrieving from trillions of tokens" by Sebastian Borgeaud et al (Dec 2021), which was referenced 5 times, including in the article My Top 5 Predictions for AI in 2022 in Towards Data Science. The paper got social media traction with 416 shares. A user, @iamtrask, tweeted "So if 96% of GPT-3 can be replaced with a database lookup, what do you think those 96% of parameters were doing in the first place? Smaller is better. Smaller is intelligence. Really great work by Borgeaud et al. Paper: Summary".
The paper shared the most on social media this week is by a team at OpenAI: "Grokking: Generalization Beyond Overfitting on Small Algorithmic Datasets" by Alethea Power et al (Jan 2022) with 167 shares. The researchers propose to study generalization of neural networks on small algorithmically generated datasets. @JMGrange (John Grange) tweeted "This paper had a couple things that totally defied my own intution. Less data = more, not less optimization🤯".
Over the past week, 18 new papers were published in "Computer Science - Multiagent Systems".
The paper discussed most in the news over the past week was "Social Neuro AI: Social Interaction as the dark matter of AI" by Samuele Bolotta et al (Dec 2021), which was referenced 5 times, including in the article Scientists say social interaction is ‘the dark matter of AI’ in The Next Web. The paper also got the most social media traction with 61 shares. On Twitter, @introspection observed "Social Neuro-AI: Social Interaction as the « dark matter » of AI — New year, new horizons. #Preprint about our future research at Kudos to for his first paper! Summary in my talk at cc &".
Over the past week, 21 new papers were published in "Computer Science - Neural and Evolutionary Computing".
This week was active for "Computer Science - Robotics", with 64 new papers.
The paper discussed most in the news over the past week was "PIEEG: Turn a Raspberry Pi into a Brain-Computer-Interface to measure biosignals" by Ildar Rakhmatulin et al (Jan 2022), which was referenced 4 times, including in the article Turning a Raspberry Pi Into a Brain-Computer Interface? Researchers Open-Source the Low-Cost, High-Precision PIEEG in SyncedReview.com. The paper got social media traction with 62 shares. The authors present an inexpensive, high - precision, but at the same time, easy - to - maintain PIEEG board to convert a RaspberryPI to a Brain - computer interface. On Twitter, @M157q_News_RSS observed "Turn a Raspberry Pi into a Brain-Computer-Interface to Measure Biosignals Article URL: Comments URL: Points: 107 # Comments: 25", while @NeuroTechZH said "Congratulations it's great work you're doing bringing EEG to the Raspberry Pi platform. #bci #iot #eeg".