Phi-3 mini model released under MIT! 🚀 Last Week Llama 3, this week Phi-3 🤯 Microsoft Phi-3 comes in 3 different sizes: mini (3.8B), small (7B) & medium (14B). Phi-3-mini was released today, claiming to match Llama 3 8B performance! 🚀

3.8B TL;DR:
2️⃣ Instruct Versions with 4k…

Cool report from Stanford AI and all significant players.

Perhaps you will net real through all 500+ pages, but interactive summary is surely worth 1 hour of your time.

I think me and my colleagues managed to produce a cool post recently ;)

Cool way to announce new models 🫣

1. как человек, который проводил алгоритмические собесы в Google, а теперь собесит сеньоров, выкатываю тред про то, как готовиться к алгоритмическим собеседованиям!

A Comprehensive Guide to Semantic Chunking for RAG 📑

A better approach to naive sentence/token chunking is to group semantically similar information, leading to better retrieval + more complete synthesis.

This post by Florian June is a comprehensive guide towards semantic…

The dashboard for the new full-stack FastAPI project template is taking shape. ✨🚀

There's a frontend dev dedicated to this. 😎

I'm so excited to share my monograph on Foundations of Vector Retrieval w/ you all. If you're a grad student considering theory-inclined research on data structures and algorithms for nearest neighbor search, I hope you find this to be a helpful resource. arxiv.org/abs/2401.09350

LLM Self-Consistency over Tabular Data 🤝📊

There are two ✌️ main ways to use LLMs to query tabular data: textual reasoning (dump table in prompt), and symbolic reasoning (SQL, Pandas/Python).

The problem ⛔️: each works well in certain settings but not others.

In a new paper,…

Share linkedin.com/pulse/how-buil…

Sad to hear...

Today, I'm sharing AgentSearch, an open-core effort to make humanity's knowledge accessible for LLM agents.

To start, I have embedded all of Wikipedia, Arxiv, filtered common crawl, and more.

The result is over 1 billion embedding vectors, try it at search.sciphi.ai.

I’m excited to share News Déjà Vu (newsdejavu.github.io), which uses a custom large language model to retrieve historical news articles that are the most similar to modern news articles. (1/4)

The cancel culture is invented by commies!

The Peril of Politicizing Science | The Journal of Physical Chemistry Letters pubs.acs.org/doi/full/10.10…

On behalf of our co-authors Tomáš Mikolov, Ilya Sutskever and Kai Chen, Greg Corrado and I were delighted to accept the #NeurIPS2023 Test of Time Award for the 'word2vec' paper (arxiv.org/abs/1310.4546). Thanks to the NeurIPS Conference test of time committee for honoring us with this…

Andrew Ng always has some of the most intuitive ways to explain AI concepts (simple or complex)

One of the greatest educators in the AI space.

How to Build and Evaluate Agents over your Data 🤖🧪

The next step beyond RAG is agents. We’re excited to feature a full workshop with DeepLearning.AI and TruEra showing you how to build an agent over your data.

In this workshop you’ll learn the core concepts around the…

That's fantastic!!!

Vector Databases: Where Geometry Meets Machine Learning thenewstack.io/vector-databas… Ivan Yamshchikov #VectorDatabases #MachineLearning #AI #Geometry

We release LeoLM, a series of German speaking chat models based on Lama 2 7b and 13b. They preserve their ability to understand english text. We also also versions that can reply in English and in German & perform translation between these languages: laion.ai/blog/leo-lm/