TLDR AI 2023-11-16

When a New York City developer purchased an untitled painting for $19,000 in 1983, he wasn’t expecting a 5,798x return on his investment. But that's exactly what happened when he sold the Basquiat work for a stunning $110,000,000 in 2017.

Now, everyday investors are unlocking access to the very same market thanks to the blue-chip art investing platform, Masterworks. When Masterworks sells a painting – like the 16 it's already sold – investors reap their portion of the profits.

Offerings can sell out in minutes, but TLDR AI readers can skip the waitlist to join with this exclusive link.

Investing involves risk and past performance is not indicative of future returns. See important Reg A disclosures and aggregate advisory performance at masterworks.com/cd

Microsoft has developed its own custom AI chip, Azure Maia AI, and an Arm-based CPU for Azure data centers, Azure Cobalt, aiming to reduce reliance on Nvidia and prepare for an AI-driven future. The chips are set to launch in 2024.

Google’s Search Generative Experience (SGE) is slowly being made available to more countries in many languages after debuting earlier this year. The company is taking a slow and measured pace to this release as it touches many of its key search systems.

GitHub's Copilot X revolutionizes software development by embedding AI in every stage. The new Copilot Chat, powered by GPT-4, enables natural language coding. It arrives in December 2023. Copilot Enterprise, tailored for organizational codebases, will launch in February 2024. Enhanced AI security features and the expanding Copilot Partner Program further enrich GitHub's offerings.

This team created a new RISC-V hardware designed for vision and trained an ultra-small YOLO computer vision model which enables real-time object detection at low power and extremely low latency.

Optimizing models according to preferences was one of the breakthroughs that enabled the widespread adoption of ChatGPT. Similar technologies can also be used to improve the correctness and factuality of models, reducing errors in medical recall by over 50%.

Developing against OpenAI APIs can be finicky and expensive. MockGPT by WireMock is a readymade mock API for ChatGPT, GPT-3.5, etc. Use it to write and test your code without fees and frustrating wait times — and switch one line of code to go live with the real API. Read the blog or use MockGPT for free

Qwen is a series of models from Alibaba Cloud. This installment is a general-purpose audio-to-language model. It can perform a number of tasks for music, speech, and more.

DeepMind has been working on weather forecasting for a few years now. Its most recent model, built on graph neural networks, is extremely accurate for 10-day forecasts. DeepMind released code and weights so you can run it locally (if you have satellite data).

This project introduces a new method called "distance indexing" to improve video frame interpolation (VFI), which traditionally struggles with predicting precise object movements. This approach clarifies learning objectives and reduces blurriness by giving models explicit information about the distance an object travels between frames.

Someone recently put Mistral’s 7B model into a stuffed parrot and it made the rounds online. This post dives into the basic ideas they used to get it working.

AI-generated music could be the future of what we consume or what helps us shape our listening habits - this could potentially be music’s “Midjourney moment”.

Those who said AI would doom us all are now having to revisit their positions. “AI doom” really only helped the big guys, and opportunities abound for the open source movements happening in AI.

Resolve 80% of your customer support questions with no code.

LLM Sherpa provides strategic APIs to accelerate large language model use cases.

Google’s deepmind research group has released a number of animations and art to build an understanding of different AI systems. The animations are somewhat odd but aesthetically beautiful.