« Chère Aurore Bergé »
Boxing Day, épisode 29
« On l’aura compris, vous êtes prête à absolument tout pour servir vos ambitions, quitte à tordre le réel, caricaturer, mentir, diffamer, calomnier et renier vos supposées «convictions» lorsque le vent tourne. »
Via @blast-info.fr
shorturl.at/uOi7K
First image: King Kong and Godzilla fighting. Godzilla is labelled "english" King Kong is labelled "arabic". Godzilla says "ninety-nine" King Kong says "nine ninety". Second image: King Kong and Godzilla fleeing from a Shiba Inu wielding a baseball bat labelled "french" saying "four twenty ten nine"
8 new types of RAG
▪️ DeepRAG
▪️ RealRAG
▪️ Chain-of-Retrieval Augmented Generation (CoRAG)
▪️ VideoRAG
▪️ CFT-RAG
▪️ Contextualized Graph RAG (CG-RAG)
▪️ GFM-RAG
▪️ URAG
Save the list and check this out for more info: huggingface.co/posts/Ksenia...
Diagram showcasing an open-source AI stack by ByteByteGo. Includes tools for frontend (Next.js, Vercel, Streamlit), embeddings/RAG libraries (Nomic, Cognita, LLMWare), backend (LangChain, Huggingface), data/retrieval (Postgres, Milvus), and large language models (Llama 3.3, Mistral, Phi).
Exploring the Open Source AI Stack! 🖥️ From Next.js to Llama 3.3, this stack covers it all: frontend, embeddings, retrieval, and large language models. A complete toolkit for building the future of AI. 🚀 #AI #OpenSource #MachineLearning #AIStack #Developers
8 new types of RAG
▪️ DeepRAG
▪️ RealRAG
▪️ Chain-of-Retrieval Augmented Generation (CoRAG)
▪️ VideoRAG
▪️ CFT-RAG
▪️ Contextualized Graph RAG (CG-RAG)
▪️ GFM-RAG
▪️ URAG
Save the list and check this out for more info: huggingface.co/posts/Ksenia...
For those saying it wasn't a Nazi salute.
Keep posting it. Keep sharing it.
Chamath Palihapitiya @chamath · 2h Wikipedia lies. Mainstream media lies. They will contort the truth to make you believe in a bogeyman. In falling for it, you’re giving up your God given ability to think for yourself.
Shaun Maguire @shaunmmaguire Wikipedia has been ideologically captured for years It's important that this is finally coming to light The most extreme bias is probably against Israel Foundation AI models are being trained on this biased data as if it's ground truth
Elon Musk @elonmusk · 3h Defund Wikipedia until balance is restored!
Elon Musk @elonmusk · 3h Since legacy media propaganda is considered a “valid” source by Wikipedia, it naturally simply becomes an extension of legacy media propaganda!
having succeeded in purchasing positions of influence inside the White House, the billionaires and tech VCs have now set their sights back on Wikipedia
DeepSeek R1 1.5B running FULLY LOCALLY in your browser , powered by WebGPU🔥
Try it out directly here: huggingface.co/spaces/webml...
Les mails pro pourris - tellement vrai 🤣🤣🤣
Small but mighty! 🚀 Phi-4 proves you don’t need billions of parameters to achieve top-tier performance. Compact, efficient, and leading the frontier for smaller models. Try it on #ollama 😍
#AI #ML #EfficientAI #Transformers #AIInnovation
Diagram showcasing an open-source AI stack by ByteByteGo. Includes tools for frontend (Next.js, Vercel, Streamlit), embeddings/RAG libraries (Nomic, Cognita, LLMWare), backend (LangChain, Huggingface), data/retrieval (Postgres, Milvus), and large language models (Llama 3.3, Mistral, Phi).
Exploring the Open Source AI Stack! 🖥️ From Next.js to Llama 3.3, this stack covers it all: frontend, embeddings, retrieval, and large language models. A complete toolkit for building the future of AI. 🚀 #AI #OpenSource #MachineLearning #AIStack #Developers
So true ! 😅
A radar chart comparing AI models (GPT-4o-202405, Gemini-1.5 Pro, Qwen2-VL 7B, GLM-4-Voice 9B, MiniCPM-o 2.6 8B) across tasks like speech conversation, visual understanding, and live streaming. GPT-4o-202405 dominates most categories with exceptional performance.
Radar of AI capabilities! 🛠️ GPT-4o-202405 leads across diverse benchmarks in speech, visual understanding, and live streaming tasks. From TextVQA to AISHELL-1, it's a game-changer. #AI #ML #GPT4 #AIInnovation #MultimodalAI
Small but mighty! 🚀 Phi-4 proves you don’t need billions of parameters to achieve top-tier performance. Compact, efficient, and leading the frontier for smaller models. Try it on #ollama 😍
#AI #ML #EfficientAI #Transformers #AIInnovation
5️⃣ Big challenges remain ⚠️
User-friendliness, interoperability, and privacy are hurdles that need to be addressed before agents can truly thrive.
An exciting leap forward, but there’s still work to do.
👉 www.kaggle.com/whitepaper-a...
What’s your take on this AI evolution? 💭
4️⃣ Building agents is getting easier 💡
Platforms like LangChain and Vertex AI make it simpler to design and deploy agents, even for complex tasks.
3️⃣ Tools make all the difference 🛠️
APIs, databases, and real-time data give agents superpowers. They’re no longer limited to answering questions—they can take action.
2️⃣ Success depends on the setup 🔧
Effective agents require three key components:
A smart model for decision-making,
Tools to interact with the real world,
A system to manage planning and execution.
1️⃣ Agents are AI on steroids 🚀
They go beyond standard models by using tools, making plans, and taking actions—similar to how humans solve problems.
🚨 Google releases a white paper on AI agents
The document dives into the basics of LLM agents and includes a quick implementation guide using LangChain. Here are the 5 key takeaways:
🚨 Microsoft launches MarkItDown: an open-source tool for converting PDFs, Word, Excel, images, and audio to Markdown—perfect for LLMs! 🎉
✨ Key features:
- Auto OCR & EXIF extraction
- LLM-powered image descriptions
- CLI, Python API, Docker
- Batch conversions
- Streamline your workflows today! 🚀
If you’re wondering how quickly AI performance is advancing, this graph says it all. The growth is practically vertical! 🚀
agree with you
Table comparing the performance of AI models (phi-4, phi-3, Qwen 2.5, GPT, Llama-3.3) on benchmarks like MMLU, GPQA, MATH, HumanEval, MGSM, SimpleQA, and DROP. Scores highlight strengths of different models in reasoning, coding, and QA tasks across small and large models.
New AI model benchmarks show phi-4 outperforming on GPQA, MATH, and reasoning tasks, while GPT-4o leads in HumanEval and MMLU. The race in AI performance intensifies across domains! 🚀 #AI #MachineLearning #Benchmarking #NLP #ArtificialIntelligence #Phi4 #GPT
Hi, awesome list! Would love to join the crew if there is room left
yep, add me please 🙏🏾
When I click on your link, I get this.
Thanks 😁, I just repost it ! 😁
thank you very much 👍
