TLDR 2024-04-10

Google's new Vids app can create collaborative shareable videos for things people do at work. The app makes making videos as easy as making slides - no video production is required. Users line up assets inside the app and edit it all into a finished video. They can choose to do all of this themselves or ask Google's Gemini AI to build storyboards, write scripts, read the scripts aloud using text-to-speech, and create images to use in the video. People with access to the video can add comments, leave notes, and make edits. Vids will launch in public beta this summer.

Intel gave the first architectural details of its Gaudi 3 third-generation AI processor at Vision 2024 this week in Phoenix, Arizona. Gaudi 3 is made up of two identical silicon dies, each with a central region of 48 megabytes of cache memory, joined by a high-bandwidth connection, surrounded by four engines for matrix multiplication and 32 programmable units called tensor processor cores. It produces double the AI compute of Gaudi 2 using 8-bit floating-point infrastructure. It also provides a fourfold boost for computations using the BFloat 16 number format.

Starfish, a neural interface company co-founded by Gabe Newell, recently updated its website, indicating that its team is working hard on its brain-computer solutions. The company is developing minimally invasive one-dimensional implants for neuromodulation and neural recording. Newell's interest in the field goes back to at least 2021, when he brought up the idea of using brain-computer interfaces for gaming. Valve has previously worked to create open source software to allow developers to understand signals coming from people's brains.

Meta plans to release an initial version of its next-generation Llama 3 large language model within the next month. The company will release a number of different models with different capabilities and versatilities during the course of the year. Llama 3 will be able to answer a wider range of questions compared to its predecessor, including questions regarding more controversial topics. Meta has not released any details about the model's size, but it is expected to have about 140 billion parameters - the biggest Llama 2 model has 70 billion.

Chronon is a platform that allows organizations to power AI/ML projects without needing to worry about orchestration by abstracting away the complexity of data computation and serving for AI/ML applications. It can perform batch and streaming computation, scalable backfills, low-latency serving, and more. Chronon can utilize all of the data within an organization.

Google's Gemini Code Assist is an enterprise-focused AI code completion and assistance tool. It was previously offered under the now-defunct Duet AI branding which became generally available in late 2023. Code Assist is both a rebrand and a major update. It uses Gemini 1.5 Pro, which has a million-token context window. Code Assist will be available through plug-ins for popular editors like VS Code.

Apple is leasing an office building in the Miami area. The 45,000 square feet of space, located at The Plaza Coral Gables, is still under construction. Apple is also planning to open a new retail store at the Miami Worldcenter in the heart of the city. Multiple tech companies have expanded to South Florida, including Amazon and Microsoft.

Beeper, the company that attempted to launch an app that let Android users use iMessage a few months ago, has been acquired by Automattic, the giant that owns WordPress. As part of the deal, Beeper is opening up its messaging app, which attempts to corral all messaging services into one inbox, to everyone across platforms, and shutting down its waitlist for good. Automattic's CEO says that messaging will be the next big pillar of the company. Its team intends to replace a lot of messaging methods with an open-source system.

This post contains a list of 130 tips and mindsets for living a better life.

GPT-4 Turbo with Vision allows developers to call on a single model for both text and image processing.

The fragile balance of open source will be disrupted if the community comes to believe that maintaining that balance is more trouble than it's worth.

It's possible to build reliable systems out of AI models by writing simple prompts, building an eval system to do prompt engineering and improve performance in a principled way, deploying AI systems with good observability, investing in Retrieval Augmented Generation (RAG), and fine-tuning models using data gathered from the process.

Google's new Arm-based CPU, Axion, will be used to support Google's AI workloads before it rolls out to Google Cloud business customers later this year.