Email Details

OpenAI raises funds at $157B valuation 💰,Chromebook AI features 💻, Embedding Knowledge Graphs 🌐

OpenAI has raised $6.6B in new funding at a $157B post-money valuation to continue scaling AI. Investors include Microsoft and Nvidia. 

TLDR

Together With

TLDR AI 2024-10-03

Nylas ExtractAI: Turn user emails into structured data (Sponsor)

The world (still) runs on email. Invoices, product orders, travel plans - they're all in the inbox. But unstructured data and wildly disparate email formats make extracting the parts you need difficult.

ExtractAI by Nylas is the new way to sync, filter, and extract user data. With just a few lines of code, you can now deploy advanced ML, NLP, and LLMs to retrieve first-party data from your users' inboxes:

  • 92% cost savings compared to other AI / LLM tools
  • 99.9% accuracy SLA for order data
  • Real-time sync with no email forwarding required

See what you can build with email data →

🚀

Headlines & Launches

OpenAI's latest funding (1 minute read)

OpenAI has raised $6.6B in new funding at a $157B post-money valuation to continue scaling AI. Investors include Microsoft and Nvidia.
Google adds a multi-functional quick insert key and new AI features to Chromebook Plus (3 minute read)

Google has introduced new Chromebook models from Samsung and Lenovo featuring AI-powered tools. Key updates include a "Help me read" tool for summarizing PDFs and web content and live translation for video calls. Enhanced AI features are also coming to existing Chromebooks, such as AI writing tools and a built-in speech-to-text recorder.
Brain-like Computers Tackle the Extreme Edge (5 minute read)

BrainChip has announced the Akida Pico, a milliWatt-level neuromorphic chip designed for AI inference in power-constrained devices like wearables. The chip mimics brain activity to minimize energy use, targeting real-time processing needs in small user devices. Akida Pico joins similar edge-oriented neuromorphic chips from companies like Innatera and SynSense.
🧠

Research & Innovation

Scaling Optimal LR Across Token Horizon (18 minute read)

This paper explores how to scale learning rate when you increase the amount of training data a model sees. Llama did exponential -0.28 and this paper suggests exponential -0.33.
Embedding Knowledge Graphs (28 minute read)

This paper introduces a new way to approach knowledge graph embedding by using group theory to incorporate uncertainty.
Person Re-ID with Textual Prompts (14 minute read)

This paper presents a new method, Prompt-guided Feature Disentangling (ProFD), for addressing occlusion issues in person Re-Identification (ReID) tasks.
🧑‍💻

Engineering & Resources

💧GPU Droplets on DigitalOcean (Sponsor)

Powerful enough for the AI serious, simple enough for the AI curious. Try GPU Droplets:  NVIDIA H100-powered virtual machines that let you spin up virtualized GPUs on-demand, and spin them down at will. With DigitalOcean, there's minimal setup and you can jump right into training and deploying models. Sign up and get started today!
Local File Organization (GitHub Repo)

This tool uses Llama 3.2 3B and Llava-1.6 to organize your files on your computer into logical sections based on content.
New Image Restoration Algorithm (2 minute read)

Posterior-Mean Rectified Flow (PMRF) is an innovative algorithm for photo-realistic image restoration.
Improve LLM Collaboration (GitHub Repo)

RouterDC is a novel method that enhances collaboration between multiple large language models (LLMs) by using query-based routing. It employs contrastive learning to select the best model for each query, improving performance over existing routing methods.
🎁

Miscellaneous

AI Can Best Google's Bot Detection System, Swiss Researchers Find (3 minute read)

Researchers from ETH Zurich used AI to achieve a 100% success rate in solving Google's reCAPTCHA v2. This breakthrough highlights AI's capability to overcome image-based CAPTCHA systems. As bots become smarter, CAPTCHA's future as a security technology remains uncertain.
Negligence Liability for AI Developers (20 minute read)

This article discusses the need for a negligence-based approach to AI accountability, focusing on the human factors behind AI systems. It critiques current regulatory frameworks that overlook the role of AI developers and suggests California's AI safety bill as a potential model. The article also explores the challenges in defining reasonable care in AI and the implications of treating AI developers as professionals.
I am tired of AI (6 minute read)

This author expresses frustration with the overuse and marketing of AI, particularly in software testing and conference proposals. They argue AI tools often prioritize speed over quality and lack the unique insight of human-generated work. Despite recognizing some beneficial uses, they criticize the proliferation of mediocre AI-generated content.
⚡

Quick Links

How Baseten built function calling + JSON mode for open-source and fine-tuned LLMs (Sponsor)

An engineering deep dive about building new capabilities into a customized version of NVIDIA's Triton inference server. Covers schemas and tools, building a state machine, and forcing valid output. Read the blog
OpenAI Training Data to Be Inspected in Authors' Copyright Cases (4 minute read)

OpenAI has agreed to allow authors to review its training data amid copyright litigation.
The Four Short Term Winners of AI (6 minute read)

The global AI arms race is dominated by Big Tech firms, chipmakers like NVIDIA, intellectual property lawyers, and The Big 4 consulting firms.
ByteDance will reportedly use Huawei chips to train a new AI model (2 minute read)

ByteDance plans to train a new AI model using Huawei's Ascend 910B chips.

Love TLDR? Tell your friends and get rewards!

Share your referral link below with friends to get free TLDR swag!
https://refer.tldr.tech/6d412934/2
Track your referrals here.

Want to advertise in TLDR? 📰

If your company is interested in reaching an audience of AI professionals and decision makers, you may want to advertise with us.

If you have any comments or feedback, just respond to this email!

Thanks for reading,
Andrew Tan & Andrew Carr


If you don't want to receive future editions of TLDR AI, please unsubscribe from TLDR AI or manage all of your TLDR newsletter subscriptions.

© 2024 Email Dashboard. All rights reserved.