Highlights in AI: is MS Marco in trouble? — October Edition

We discuss some recent highlights in the world of Artificial Intelligence: from news to code and academic research. Pupil Shapes Reveal GAN-generated Faces, Multimodal Prompt Engineering, CNNs vs Transformers vs MLPs for vision, Primer Evolved Transformer, FLAN, plus the pressing question of whether MS MARCO has reached end of life in neural retrieval evaluation. This month, besides the regulars Jakub Zavrel, Marzieh Fadaee and Sergi Castella i Sapé, we have two special invited panelists: Leo Boytsov from Bosch AI and Negar Arabzadeh, Information Retrieval PhD student at the University of Waterloo. Timestamps: 00:00 Introduction 4:05 What we do at Zeta Alpha 11:27 RecSys 2021 Conference highlights 13:29 LayoutParser: Deep Learning based document parsing 14:32 Huggingface infinity, accelerated inference with Graphcore and specialized hardware for ML 18:27 A Battle of Architectures: CNNs vs. MLPs vs. Transformers 21:23 Finetuned Language Models are Zero-shot Learners 2