Issue 76 5th November 2024

Highlights

Evaluating generative AI: Best practices for developers

How do you ensure your app's outputs consistently meet your quality standards and business requirements? Enter the essential world of generative AI..

Hidden traces of humanity: what AI images reveal about our world – podcast

As generative AI advances, it is easy to see it as yet another area where machines are taking over – but humans remain at the centre of AI art, just in ways we might not expect. By Rachel Ossip.

Anthropic hikes the price of its Haiku model

Anthropic’s newest AI model has arrived. But it’s pricier than the last generation, and unlike Anthropic’s other models, it can’t analyze images just yet. Claude 3.5 Haiku, which Anthropic announced last month, matches or bests the performance of Anthropic’s once-flagship, Claude 3 Opus, on specific benchmarks. Available through Anthropic’s API and a number of third-party […]

Research

Differentiable GPU-Parallelized Task and Motion Planning

Video

Episode 309 - Azure AI Search with Farzad Sunavala

Join us in this episode where we talk with Farzad Sunavala about Azure AI Search.

NVIDIA’s New Ray Tracing Tech Should Be Impossible!

Articles

Google and Meta could face defamation risks over AI-generated responses, Australian experts warn

Lawyer says tech platforms could be held liable for what their AI ‘spits out’, as Google Maps rolls out new features with Gemini Follow our Australia news live blog for latest updatesGet our breaking news email, free app or daily news podcastMeta and Google using user comments or reviews as part of generative AI responses to queries on restaurants or to summarise sentiment could introduce new defamation risks, experts have warned.

A Simple LLM Agent Deployment Tutorial

Many tutorials show how to implement an LLM agent. However, resources on deploying these agents behind an API or a user-friendly UI are limited. This post addresses this gap with a step-by-step guide to implementing and deploying a minimal yet functional LLM agent. This provides a starting point for your LLM agent proof of concept, whether for personal use or to share with others.

AI-powered microgrids facilitate energy resilience and equity in regional communities

When augmented with AI, small power grids can create opportunities for decentralized, equitable, and resilient power. Microsoft’s collaboration shows AI’s potential to optimize energy distribution and empower communities.

The post AI-powered microgrids facilitate energy resilience and equity in regional communities appeared first on Microsoft Research.

Build a Conversational AI Assistant with the Azure OpenAI Assistants API

The Assistants API orchestrates tool usage, calling database queries using function calling, the code interpreter for in-depth analysis and visualizations, and file search to provide additional context to the language model using provided documents.

How to Create a RAG Evaluation Dataset From Documents

Automatically create domain-specific datasets in any language using LLMs.

New OpenAI feature: Predicted Outputs

Interesting new ability of the OpenAI API - the first time I've seen this from any vendor. If you know your prompt is mostly going to return the same content - you're requesting an edit to some existing code, for example - you can now send that content as a "prediction" and have GPT-4o or GPT-4o mini use that to accelerate the returned result.

GraphRAG in Action: From Commercial Contracts to a Dynamic Q&A Agent

In this blog post, we introduce an approach that leverages a Graph Retrieval Augmented Generation (GraphRAG) method — to streamline the process of ingesting commercial contract data and building a Q&A Agent.

Deploying LLMs with TorchServe + vLLM

The vLLM engine is currently one of the top-performing ways to execute large language models (LLM). It provides the vllm serve command as an easy option to deploy a model on a single machine. While this is convenient, to serve these LLMs in production and at scale some advanced features are necessary.

MIT Schwarzman College of Computing launches postdoctoral program to advance AI across disciplines

The new Tayebati Postdoctoral Fellowship Program will support leading postdocs to bring cutting-edge AI to bear on research in scientific discovery or music.

Upcoming events

Microsoft Ignite | November 18-22, 2024 | Chicago, IL

Join us at Microsoft Ignite on November 18-22, 2024 in Chicago, IL and online to achieve more with the power of the Microsoft Cloud, Microsoft Copilot, and AI.

Code

Claude Token Counter

Anthropic released a token counting API for Claude a few days ago. I built this tool for running prompts, images and PDFs against that API to count the tokens in them.