OpenAI expands its custom model training program (4 minute read)
OpenAI's Custom Model training program is expanding to help enterprise customers develop tailored generative AI models.
Scientists are deep-freezing corals to repopulate the ocean (9 minute read)
Coral reefs are degrading at an unprecedented rate due to pollution, overfishing, and destructive forestry and mining practices on land. Most corals could go extinct within a few generations. The Smithsonian’s National Zoo and Conservation Biology Institute's Reef Recovery Initiative aims to help save coral reefs using cryopreservation, a novel approach that involves storing and cooling coral sperm and larvae at very low temperatures and holding them in government biorepositories. The frozen assets can be used to help reseed the oceans and restore living reefs even 100 years from now. Cryopreserved sperm has already been used to produce new coral across the Caribbean.
How Stack Overflow replaced Experts Exchange (19 minute read)
This article tells the story of how and why Stack Overflow was created and discusses whether the site will have a place in engineering culture over the next decade. Before Stack Overflow, there was Experts Exchange, one of the first question-and-answer sites on the internet, coming into existence in 1996. Experts Exchange was free for over a decade, but it moved to a subscription-based model in 2004. This paywall created the perfect moment for a new site to counter-position itself, so the creators of Stack Overflow decided to create a more user-friendly version of Experts Exchange. Stack Overflow launched in 2008, the same year as GitHub, Apple's App Store, and Airbnb. The future existence of Stack Overflow will depend entirely on how well it can adapt to AI.
How Tech Giants Cut Corners to Harvest Data for A.I. (22 minute read)
OpenAI solved its data supply problem in late 2021 by developing a speech recognition tool called Whisper to transcribe audio from YouTube videos, yielding new text to train its AI systems. Its team ultimately transcribed more than one million hours of video to train GPT-4. Google and Meta have also similarly cut corners in the race to lead AI, obtaining data in legally and ethically dubious ways. This article discusses the various methods companies have used to obtain data - or have considered using to obtain data - for AI training purposes.
Opera browser dev branch rolls out support for running LLMs locally (3 minute read)
Opera's Opera One Developer browser can now run 150 different large language models from 50 different large language model families without requiring an internet connection except to download the models.
"If this one guy got hit by a bus, the world's software would fall apart." (7 minute read)
The entirety of modern computing infrastructure is built on top of thousands of projects that are built, maintained, and run entirely by one person or by extremely small teams that are more often than not unpaid volunteers.
Google Launches Android Find My Device Network (2 minute read)
Google has launched the Find My Device network for Android-based products. The Android Find My Device network can use the millions of existing Android devices to track down lost, stolen, and missing Android products. The network uses Bluetooth and works even when Android devices are offline and do not have a cellular or Wi-Fi connection. The network will also work with third-party Bluetooth trackers starting in May.
Notes on how to use LLMs in your product (17 minute read)
It's easy to make an argument that large language models can improve internal efficiency in companies somehow. It's much harder to describe a believable way for them to make products more useful to customers. This article discusses ways to meaningfully integrate large language models into existing products. It covers how to think about the technology, revamping workflows, different ways to implement models, and more. The article also covers potential issues such as copyright law, data processing agreements, and provider availability.
With Vids, Google thinks it has the next big productivity tool for work (4 minute read)
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.
Meta confirms that its Llama 3 open source LLM is coming in the next month (3 minute read)
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.
Google launches Code Assist, its latest challenger to GitHub’s Copilot (5 minute read)
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.
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.
OpenAI makes GPT-4 Turbo with Vision available to developers to unlock new AI apps (2 minute read)
GPT-4 Turbo with Vision allows developers to call on a single model for both text and image processing.
Google has made a handful of editing features previously exclusive to Pixel devices and paid subscribers free to all Google Photos users. The features include the AI-powered Magic Editor, Magic Eraser, Photo Unblur, and Portrait Light. They will only work on devices running Android 8.0 or iOS 15 or higher and on Chromebook Plus devices with at least 3GB RAM and ChromeOS version 118+. The tools will start rolling out on May 15 - it may take weeks for them to make it to all users.
Hopes rise for mRNA cancer vaccine after Moderna trial shows promise (1 minute read)
Moderna's mRNA cancer vaccine, developed to target melanoma, might also treat a form of head and neck cancer. Data from an early trial showed a greater survival rate for patients who took Moderna's cancer vaccine alongside immunotherapy treatments. The news sent Moderna's share price soaring.