On evaluating LLMs: Let the errors emerge from the data | AI & ML Monthly

Video Description

Welcome to machine learning & AI monthly for May 2025. This is the video version of the newsletter I write every month which covers the latest and greatest (but not always the latest) in the world of AI and ML. Read the issues online: - AI/ML Monthly May 2025 (this video) — https://zerotomastery.io/blog/ai-and-machine-learning-monthly-newsletter-may-2025/ - AI/ML Monthly April 2025 — https://zerotomastery.io/blog/ai-and-machine-learning-monthly-newsletter-april-2025/ - AI/ML Monthly March 2025 — https://zerotomastery.io/blog/ai-and-machine-learning-monthly-newsletter-march-2025/ My links: Download Nutrify (my startup) - https://nutrify.app Download KeepTrack (my other startup) - https://keeptrack.app Learn Hugging Face - https://dbourke.link/ZTM-HF-Text-Classification Learn AI/ML (beginner-friendly course) - https://dbourke.link/ZTMMLcourse Learn TensorFlow - https://dbourke.link/ZTMTFcourse Learn PyTorch - https://dbourke.link/ZTMPyTorch My ML blog - https://learnml.io Read my novel Charlie Walks - https://www.charliewalks.com Personal website - https://www.mrdbourke.com Timestamps: 00:00 - Welcome - https://zerotomastery.io/blog/ai-and-machine-learning-monthly-newsletter-may-2025/ 00:18 - Apache 2.0 object detection models: https://www.learnml.io/posts/apache-object-detection-models/ 02:17 - Netflix's Foundation Model for Recommendations: https://netflixtechblog.com/foundation-model-for-personalized-recommendation-1a0bd8e02d39 06:32 - AI Workflows vs AI Agent Patterns: https://www.philschmid.de/agentic-pattern 11:19 - Spotify's Text2Tracks System: https://research.atspotify.com/2025/04/text2tracks-improving-prompt-based-music-recommendations-with-generative-retrieval/ 16:31 - LinkedIn's JUDE Job Embeddings: https://www.linkedin.com/blog/engineering/ai/jude-llm-based-representation-learning-for-linkedin-job-recommendations 21:10 - Mixedbread's OCR Ceiling and RAG Performance: https://www.mixedbread.com/blog/the-hidden-ceiling 25:39 - NVIDIA NeMo Hugging Face Support: https://developer.nvidia.com/blog/run-hugging-face-models-instantly-with-day-0-support-from-nvidia-nemo-framework/ 26:30 - NVIDIA Kaggle Competition Tips: https://developer.nvidia.com/blog/grandmaster-pro-tip-winning-first-place-in-a-kaggle-competition-with-stacking-using-cuml/ 28:21 - Benedict Evans on AI Adoption: https://www.ben-evans.com/presentations 30:14 - Analyzing LLM Application Failures: https://mlops.systems/posts/2025-05-23-error-analysis-to-find-failure-modes.html 35:19 - FastVLM: Efficient Vision Encoding: https://github.com/apple/ml-fastvlm 35:44 - nanoVLM: 750-Line VLM Pipeline: https://github.com/huggingface/nanoVLM 37:57 - SmolVLM demo in the browser: https://github.com/ngxson/smolvlm-realtime-webcam 38:26 - D-FINE Object Detection Models: https://huggingface.co/collections/ustc-community/d-fine-68109b427cbe6ee36b4e7352 39:07 - OpenVision: 20+ Open-Source CLIP Models: https://ucsc-vlaa.github.io/OpenVision/ 40:46 - EdgeTAM: Track Anything in Video: https://github.com/facebookresearch/EdgeTAM 42:56 - Roboflow's Object Tracking Library: https://github.com/roboflow/trackers 44:00 - Osmosis: Structured Data Extraction: https://huggingface.co/osmosis-ai/Osmosis-Structure-0.6B 45:06 - NVIDIA Parakeetv2: https://huggingface.co/nvidia/parakeet-tdt-0.6b-v2 47:19 - LLMDet: Open-Vocabulary Object Detection: https://huggingface.co/fushh7/LLMDet 48:56 - DeepSeek release two new reasoning models: https://huggingface.co/deepseek-ai/DeepSeek-R1-0528 50:42 - Xiaomi's MiMo LLM and VLMs: https://huggingface.co/collections/XiaomiMiMo/mimo-6811688ee20ba7d0682f5cb9 52:29 - ByteDance's Dolphin Document Parser: https://github.com/bytedance/Dolphin 54:47 - CountGD: https://www.robots.ox.ac.uk/~vgg/research/countgd/ 54:49 - Papers section: https://www.arxiv.org/abs/2412.13303 1:00:00 - Company releases: https://blog.google/technology/developers/google-ai-developer-updates-io-2025/ 01:03:16 - Media, podcasts and more: https://youtube.com/playlist?list=PLlrxD0HtieHgKcRjd5-8DT9TbwdlDO-OC&si=anCq_0bzf4zrcFN6