TLDR AI 2023-09-26

Amazon invests $4B in Anthropic 💰, ChatGPT can see & hear 👀, business document reading 📑

The window to invest in this exciting Smart Home startup is open for a limited time! (Sponsor)

Ring 一 Acquired by Amazon for $1.2B

Nest 一 Acquired by Google for $3.2B

➡️ If you missed out on these spectacular early investments in the Smart Home space, here’s your chance to grab hold of the next one.

RYSE is the tech firm that’s poised to dominate the Smart Shades market (growing at an astonishing 55% annually), and their public $1.25/share round is open for a limited time!

They’ve just secured their first retail distribution deal, and you can invest in their company before their product stocks the shelves of over 100 Best Buy stores this year.

⌛Don't miss out and invest now!

Headlines & Launches

ChatGPT can see, hear, and speak now (8 minute read)

ChatGPT is getting an upgrade in modality with a number of additions. You will soon be able to speak to the model and have it speak back. Users will also be able to upload images and ask questions about them.

Spotify Is Going To Clone And Translate Podcasters’ Voices (1 minute read)

Spotify has introduced an AI-powered voice translation feature that can reproduce podcasts in other languages using the original podcaster's voice. The technology relies on OpenAI's Whisper for transcription and likely for voice replication, allowing translations to maintain the podcaster's unique voice.

Amazon invests up to $4B in Anthropic (3 minute read)

Amazon is investing up to $4B in Anthropic with a set of trades in IP and compute options on the table. Anthropic’s safety governance remains unchanged.
Research & Innovation

Advanced Scene Annotation Using 3D Tools (2 minute read)

NeuralLabeling is a tool for detailed scene annotations using 3D techniques.

A Model That Understands and Labels 3D Scenes Using Text Descriptions (12 minute read)

Researchers have created a new model called PVLFF that can look at 3D scenes and not only identify what's in them, but also distinguish between different instances of the same object—all by using text descriptions.

Improving Business Document Reading (24 minute read)

LayoutLM is a model that is good at reading business documents. It was made even smarter with two new training exercises—one to help it understand complicated layouts and another to get numeric values right.
Engineering & Resources

Odin (GitHub Repo)

Generate knowledge graphs from your data with GPT-4.

Enhancing Depth Estimation with Passive Cameras (6 minute read)

This study introduces a method that simulates active stereo in passive cameras without a physical pattern projector. By virtually overlaying patterns using depth data from external sensors, the researchers overcome the limitations of traditional setups. This approach enhances both stereo algorithms and deep learning networks on various datasets.

DeepEval (GitHub Repo)

DeepEval provides a Pythonic way to run offline evaluations on your LLM pipelines so you can launch comfortably into production.

A boy saw 17 doctors over 3 years for chronic pain. ChatGPT found the diagnosis (14 minute read)

After three years and 17 doctors, a mother used ChatGPT to aid in diagnosing her son's perplexing symptoms. The AI platform suggested tethered cord syndrome, a condition associated with spina bifida. While ChatGPT proved valuable, experts urge caution, noting AI's potential risks and limitations in medical applications.

Language Model UXes in 2027 (13 minute read)

This article explores what large language model-powered software might look like in the medium-term future. LLMs are going to significantly shift how we interact with computers. The current state of the technology seems rudimentary and it's increasingly clear that generative AI can't quite replace the other UX innovations that have been introduced over the last several decades. LLM UXes will likely evolve to have a memory of past interactions in order to provide more relevant responses.

Shortwave launches AI email assistant to rival Google’s Duet AI (5 minute read)

Shortwave claims to have launched the most capable AI assistant for email on the planet. Notably, its assistant has the ability to search your entire email history using embedding-based search, so it can answer questions about anything you've sent or received, not just specific emails.
Quick Links

Capsule (Product)

The AI-powered video editor for content and marketing teams.

FireCut AI (Product Launch)

FireCut (for Adobe Premiere Pro) is an AI video editor that speeds up your editing by automating time-consuming tasks.

New Forethought tool lets you build workflows with natural language (2 minute read)

Forethought has launched Autoflows, AI-driven tasks based on natural language prompts. Unlike traditional workflows, Autoflows enhances resolution rates using SupportGPT and OpenAI models.
TLDR is a daily newsletter with links and TLDRs of the most interesting stories in startups 🚀, tech 📱, and programming 💻!
Join 500,000 readers for one daily email