Email Details

James Cameron joins Stability’s board🎬, updated Gemini models🤖, Medication Manufacturing with AI💊

Decorated filmmaker James Cameron has joined the board of generative media company Stability AI to guide its transition into visual storytelling. 

TLDR

Together With

TLDR AI 2024-09-25

The missing piece in your investment portfolio? Why you should be looking at private credit (Sponsor)

Private credit—i.e., loans and debt financing from non-bank lenders—is a resilient asset class that's averaged a 12% return globally (and 14.5% on Percent). It's been popular with institutional investors for a while, especially as it has outperformed both stocks and bonds in the last 3 market downturns.

Percent simplifies access to private credit investing with low minimums (just $500) and shorter durations (9 months average) through their marketplace. Deals are available to individual accredited investors, making it one of the few places an everyday investor could get in.

Visit Percent to learn more - plus get up to $500 with your first investment.

🚀

Headlines & Launches

James Cameron joins Stability's board (8 minute read)

Decorated filmmaker James Cameron has joined the board of generative media company Stability AI to guide its transition into visual storytelling.
Updated Gemini models, reduced 1.5 Pro pricing, increased rate limits (12 minute read)

The Gemini models from Google have had a huge reduction in cost, an increase to 2 million tokens in context, and improvements to overall performance. One interesting tidbit is the step change in cost after 128k tokens.
OpenAI staffers reportedly 'taken aback' by 'ominous' logo rebranding (2 minute read)

OpenAI plans to rebrand in 2024, introducing a new logo that employees found lacking in creativity. The company is shifting from a non-profit to a for-profit structure. This rebranding aims to solidify its identity as OpenAI becomes more recognized.
🧠

Research & Innovation

Medication Manufacturing with AI (4 minute read)

MIT researchers speed up a novel AI-based estimator for medication manufacturing by 60 times.
Clothing Image Editing with Multimodal Inpainting (16 minute read)

ControlEdit is a new method for precise multimodal clothing image editing. It allows for localized edits on clothing images while maintaining consistent style and natural transitions.
Element Order Affects Agent Performance (17 minute read)

The order in which UI elements are presented significantly impacts agent performance in virtual environments. Randomizing element order reduces performance as much as removing all visible text.
🧑‍💻

Engineering & Resources

How dogfooding helped Nebius AI build its AI-centric cloud (Sponsor)

Infrastructure — and especially GPUs — is a major limiting factor in ML and AI development. Nebius AI is on a journey to build the ultimate cloud for AI developers, including top-notch NVIDIA GPUs, integrated monitoring, and access to popular MLOps tools. The secret ingredient? Feedback from their in-house LLM R&D team. Read the blog post
Video Saliency Prediction Challenge (GitHub Repo)

The AIM 2024 Video Saliency Prediction Challenge tasked participants with predicting saliency maps for a set of video sequences using a newly collected AViMoS dataset of 1,500 videos.
Dynamic Mesh Reconstruction (GitHub Repo)

Dynamic 2D Gaussians (D-2DGS) is a new technique for reconstructing accurate meshes from sparse image input. Unlike previous methods that struggle with mesh quality, D-2DGS uses 2D Gaussians to represent geometry and captures deformations through controlled points.
Train GNNs on Large Graphs with FastGL (22 minute read)

FastGL is a GPU-efficient framework designed to speed up the training of Graph Neural Networks (GNNs) on massive graphs. It reduces data traffic and enhances memory efficiency by optimizing the sampling, memory, and computation phases of GNN training.
🎁

Miscellaneous

Visualizing Piecewise Linear Networks (3 minute read)

A great post by quant firm Jane Street that explores ways to visualize networks that are piecewise linear.
DreamHoi: A Novel AI Approach for Realistic 3D Human-Object Interaction Generation Using Textual Descriptions and Diffusion Models (7 minute read)

DreamHoi has introduced a new AI technique for generating realistic 3D human-object interactions using textual descriptions and advanced diffusion models. This approach aims to bridge the gap between textual input and detailed 3D output, enhancing virtual experiences.
Apple Intelligence will support German, Italian, Korean, Portuguese, and Vietnamese in 2025 (1 minute read)

Apple's generative AI will expand to more languages, including German and Korean, by 2025. It will initially launch in American English with iOS 18.1 but will face regulatory hurdles in the EU and China. Apple is in talks with the European Commission and Chinese regulators to address these issues.
⚡

Quick Links

Salesforce Ventures ups its AI fund to $1B, doubling it again (1 minute read)

Salesforce Ventures has announced a new $500 million fund, bringing its AI investment total to $1 billion.
LinkedIn scraped user data for training before updating its terms of service (4 minute read)

LinkedIn allegedly trained AI models on user data without initially updating its privacy policy.
T-Mobile partners with OpenAI for customer service revamp (2 minute read)

T-Mobile and OpenAI will build a new customer service system powered by AI agents as part of a multi-year partnership.

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.