John Kaufhold’s Post

FLY.HEALTH has contributed 5 BIO BERT finetuned model models trained to identify medical questions and no medical questions. The models are also highly biased against PII in the question although I would not call them PII filters. We have 5 more coming, and more and more as we add languages. https://lnkd.in/dVR9iDHm Hugging Face #AI #realhealthcareai #risk #filters #applications #healthcare

Access to 1M different models is amazing! Being able to use them in production ready environments at scale confidently, without hiring dedicated AIOps teams. Priceless. That's what we at Nutanix enable, through our partnership with Hugging Face and Nutanix Enterprise AI. #AIonNutanix #GenAI

#thefutureisnow For some ChatGPT is still that new AI that can write essays and somehow everyone wants to have in their company. unbeknownst to many however there are 1 million other AI models readily available for use on Hugging Face . 1 million! looking out the window can you imagine 1 million software models that somehow magically can do things "normal" algorithms can't. Not all of these models feature such a user friendly interface such as Llama or GPT, but they all have their individual function and use. so.... there are more AI models than there are people living in Stuttgart (632k downtown, wikipedia). if you wanna count them, one model one second, please be ready to count 11 days 24/7 straight through. and it's just 2024 with ChatGPT being in the market since barely two years. think pennies, get pennies, think dollars, get dollars. so.... where will we be next year this time?

Generative AI Open source AI models are transforming workflows across various industries, offering innovative solutions without breaking the bank. By leveraging existing models, you can automate and solve specific challenges efficiently. No need to reinvent the wheel—just scan, adapt, and implement! For example: Highly Regulated Industry: Facing data privacy or sovereign data challenges? Open source solutions can help you navigate these issues effectively. Many of these tools are versatile enough to run on platforms like Arduino or even your cellphone. For example: Content Creation: Companies use Hugging Face’s GPT models to generate engaging marketing content and automate customer support responses. Healthcare: Researchers utilize Hugging Face’s BERT models to analyze medical records and improve patient care through predictive analytics2. #GenerativeAI #Innovation #Automation #TechSolutions #HuggingFace

Congratulations to the Hugging Face team for reaching 1M free public models - and to Upstage AI for achieving the #1 spot on their leader board! This is a tremendous milestone for both, and marks a pivotal moment for GenAI innovation. Hugging Face's open-source repository enables collaboration, cost savings and access to purpose built models of all sizes and helps tackle real-world business challenges - standing in sharp contrast to sprawling ‘LLM Generalists’ - which have yet to systematically address industry and vertical business problems at scale, and cost effectiveness. Last week, Ken Taylor and I spent two days in a deep dive leadership strategy session with Kasey Kyungsil Roh & the team at Upstage AI. We left deeply impressed by their models, solutions and vision. Upstage AI provides highly compelling industry-specific solutions for Cloud, On-Prem, Hybrid Cloud (and even offline available) implementations, and use cases. When your enterprise is searching for innovative solutions involving GenAI, I highly recommend that you contact the Upstage Team, and of course, always architect on AWS ;) Ken Taylor Sung Kim Clem Delangue Chris Barnes Hiroyuki (Hugh) Matsushita

Indeed "one model to rule them all" is a dangerous fallacy for any ecosystem: natural or artificial. If one or very few (model, species, thoughts, etc.) dominate any ecosystem, it will be on the path to self-extinction - it does not matter if it is a coral reef or AI ecosystem.

We just crossed 1,000,000 free public models on Hugging Face! That’s the ones the media covers like Llama, Gemma, Phi, Flux, Mistral, Phi, Starcoder, Qwen, Stable diffusion, Grok, Whisper, Olmo, Command, Zephyr, OpenELM, Jamba, Yi but also 999,984 others. Why? Because contrary to the “1 model to rule them all” fallacy, smaller specialized customized optimized models for your use-case, your domain, your language, your hardware - and generally your constraints - are better. As a matter of fact, something that few people realize is that there are almost as many models on Hugging Face that are private only to one organization - for companies to build AI privately, specifically for their use-cases. Today a new repository (model, dataset or space) is created every 10 seconds on HF. Ultimately, there’s going to be as many models as code repositories and we’ll be here for it! Cheers to the community!

Llama 3 8B beats other open source models like Mistral’s Mistral 7B and Google’s Gemma 7B, both of which contain 7 billion parameters, on at least nine benchmarks (MMLU, ARC, DROP, GPQA, HumanEval, GSM-8K, MATH, AGIEval, and BIG-Bench Hard). Whereas Llama 3 70B beats Gemini 1.5 Pro on MMLU, HumanEval and GSM-8K, and — while it doesn’t rival Anthropic’s most performant model, Claude 3 Opus — Llama 3 70B scores better than the weakest model in the Claude 3 series, Claude 3 Sonnet, on five benchmarks (MMLU, GPQA, HumanEval, GSM-8K and MATH).

Hugging Face Researchers Introduce Idefics2: A Powerful 8B Vision-Language Model Elevating Multimodal AI Through Advanced OCR and Native Resolution Techniques Quick read: https://lnkd.in/gHWWRkFr HF Page: https://lnkd.in/gJaAGSkd

Hugging Face has published a new Open LLM Leaderboard with interesting results. The new leaderboard addresses the limitations of the previous benchmarks, which had become too easy and contaminated. Meaning: some LLMs were specifically trained to perform well. Which they did, but then it didn't tell much about their general capabilities. The new leaderboard includes six high-quality, uncontaminated benchmarks covering knowledge testing, reasoning, mathematical abilities, and instruction following. These changes aim to provide more challenging and fair evaluations, ensuring that model rankings reflect genuine performance improvements. The updated leaderboard has already shown significant results, particularly for Chinese models, which went up quite a bit in the ratings. 5 out of the top 10 are Chinese models, fascinating! #generativeai #llm #opensource https://lnkd.in/dK_KqNUV