Tunbridge Wells feels like a post-apocalyptic scene right now with no tap water for 5 days now and no end in sight. Rations of water being distributed, caps on how many packs of water you can buy for a family at shops, restaurants, coffee shops, gyms, hotels all closed. Have

There are 2 career paths in AI right now: The API Caller: Knows how to use an API. (Low leverage, first to be automated, $150k salary). The Architect: Knows how to build the API. (High leverage, builds the tools, $500k+ salary). Bootcamps train you to be an API Caller. This

I will be speaking at DataFest Tbilisi in Georgia later this month. On the first day - my talk will focus on the anatomy of a production-ready multi-agent architecture focusing on orchestration, governance, and scale. On the second day I will demonstrate how to deploy a

Pushed new code to integrate with AI Foundry Agent Service using Voice Live API and a new UI component to allow easier triggering and live transcription monitoring: https://t.co/pTQ0WYJx1V [branch: feature-ai-voice-live-api]

Great overview of VaultGemma's differential privacy implementation during pre-training stage.

AFD issues - just before my demo day :| #azure

What’s your telltale sign that a piece of text was written by AI? For me it's the short 3-4 word sentences, split into multiple paragraphs to look like a poem. Each paragraph has a set of questions with follow up "profound" answers. And the biggest of all - symmetry. Too much

Celebrating Nov 28th 2024 - Alb independence day with a deep dive into Foundation Models and in particular Large Language Models

Excited to release new repo: nanochat! (it's among the most unhinged I've written). Unlike my earlier similar repo nanoGPT which only covered pretraining, nanochat is a minimal, from scratch, full-stack training/inference pipeline of a simple ChatGPT clone in a single,

The decline of a great nation officially started today unless a future government with a new leader decides to remove this abomination of a policy.

Living in the Azure, OpenAI, AWS, GCP world can sometimes feel isolated from what's happening out there in AI open source world. What are people using nowadays for LLMs / Vector stores / Orchestration frameworks etc outside of the cloud vendor offerings?

What a fascinating episode from @MLStreetTalk on Embodied AI with Noumenal's Dr. Maxwell Ramstead and Jason Fox. They describe a different paradigm from today's transformer-based AI systems, one that moves from fractured representations or 'knowledge' stored in billions of

Document Chunking is about breaking content into smaller, manageable pieces for better retrieval while preserving context. Here are a few techniques to consider: 1. Token or Character based Chunking - Only use when structure or layout of a document is not important 2.

Let's do some RAG together: The six underlying pillars of RAG are: - Document Chunking - Chunk Enrichment - Chunk Embedding and indexing - Retrieval - Results Re-ranking - Response Generation

How do people find the courage to write and publish hardcover technical books on AI these days? Given how fast AI tech is changing, doesn't any content become obsolete by the time the book is published?

Is it me or are contact center human agents trying to sound like AIs these days?

I'm not really feeling the new Netflix UI, and X just got ten times more entertaining with the new @elonmusk / @realDonaldTrump saga. 🍿🍿🍿

Grok on Azure 🙂

Everyone's having a go now at transformers not being IT for AGI. Fair. Same people asking for a paradigm shift. - Easy! Yet to see a strong proposal for what that new paradigm shift should be. - Hard! <--- Focus should be here, not verbal bashing LLMs.