ML Digest 2022-06

ML

• Caches Considered Harmful for Machine Learning - https://petewarden.com/2022/06/02/caches-considered-harmful-for-machine-learning/
• Investors pull back on artificial intelligence:

"Funding for AI-focused health startups fell 32% in Q1 2022, after nine straight quarters of steady growth, according to a fresh analysis from CB Insights."

• Hugging Face reaches $2 billion valuation to build the GitHub of machine learning - https://techcrunch.com/2022/05/09/hugging-face-reaches-2-billion-valuation-to-build-the-github-of-machine-learning/
• Flash attention is all you need (for huge networks, lol) - https://habr.com/ru/post/669506/
• FasterTransformer - https://github.com/NVIDIA/FasterTransformer/
• Spancat: a new approach for span labeling - https://explosion.ai/blog/spancat
• LinkBERT: Improving Language Model Training with Document Link - https://ai.stanford.edu/blog/linkbert/
• Imagen unprecedented photorealism × deep level of language understanding - https://imagen.research.google/
• Большая версия ruDALL-E, или Как отличить Кандинского от Малевича - https://habr.com/ru/company/sberbank/blog/671210/
• End-to-end Generative Pre-training for Multimodal Video Captioning - https://ai.googleblog.com/2022/06/end-to-end-generative-pre-training-for.html
• Gradient Update #26: Facial Recognition in the Real World and Large Language Model Advances - https://thegradientpub.substack.com/p/gradient-update-26-facial-recognition?s=r
• Gradient Update #27: Face Search Engine or Stalkerware? And a New LLM Benchmark - https://thegradientpub.substack.com/p/gradient-update-27-face-search-engine
• Last Week in AI #172: Controversy over Google's "sentient" chatbot, DALL-E Mini goes viral, Reddit bans deepfakes sub, AI to improve video calls, and more! - https://lastweekin.ai/p/last-week-in-ai-172-controversy-over
• An Illustrated Tour of Applying BERT to Speech Data - https://thegradient.pub/an-illustrated-tour-of-applying-bert-to-speech-data/
• FIGS: Attaining XGBoost-level performance with the interpretability and speed of CART - https://bairblog.github.io//2022/06/30/figs/
• Яндекс выложил YaLM 100B — сейчас это крупнейшая GPT-подобная нейросеть в свободном доступе. Вот как удалось её обучить - https://habr.com/ru/company/yandex/blog/672396/
• Making Deep Learning Go Brrrr From First Principles - https://horace.io/brrr_intro.html
• Techniques for Training Large Neural Networks - https://openai.com/blog/techniques-for-training-large-neural-networks/
• LaMDA’s Sentience is Nonsense - Here’s Why - https://lastweekin.ai/p/lamdas-sentience-is-nonsense-heres
• MODEL ENSEMBLING in PyTorch - https://pytorch.org/functorch/stable/notebooks/ensembling.html
• DALL·E mini:
• https://www.craiyon.com/
• https://wandb.ai/dalle-mini/dalle-mini/reports/DALL-E-Mini-Explained-with-Demo--Vmlldzo4NjIxODA

____________________________________________________________________________

Originally posted on - https://t.me/snakers4