Good Morning! Google announced major upgrades for game developers on Google Play, including new subscription perks and enhanced game listings. Microsoft released an Agility SDK with cutting-edge graphics features like Work Graphs for more efficient GPU utilization. OpenAI introduced the Transformer Debugger, an open-source tool to help interpret the behavior of large language models.

— Forrest Knight & Meghanadh Vasireddy

Google Reveals Major Gaming Upgrades for Developers

At their Google for Games Developer Summit, Google announced major upgrades aimed at improving gaming experiences on Google Play for both developers and players.

New Play Pass Perks: Play Pass subscribers across 21 countries can now get exclusive in-game items and discounts for popular titles like EA Sports FC Mobile, Mobile Legends: Bang Bang, Monopoly GO!, and Roblox. This adds value to the Play Pass subscription without increasing costs.

Enhanced Game Listings:

• Game store listings can now display developer YouTube videos, AI-generated FAQs, and other content.

• Helps highlight updates and keeps existing players engaged with their games.

A new interactive ad format allows ads to seamlessly blend into the game environment itself, avoiding disruptive pop-ups or video ads that interrupt gameplay.

Developers can now more easily integrate multiple sign-in options through Play Games Services. This allows syncing game progress and achievements seamlessly across devices for players.

Native PC Game Support:

• Google Play Games is adding support to publish and discover native PC game titles optimized for that platform.

• Popular games like Lineage2M, Odin: Valhalla Rising, Genshin Impact are already available.

• Allows optimized PC games to reach a wider gaming audience through Google Play.

These additions provide new tools to increase user engagement, developer revenue and multi-platform reach. Gaming remains a core focus area where Google continues investing heavily, with the goal of making Google Play the premier destination for mobile and PC game developers.

Read More Here

Microsoft’s latest Agility SDK packs cutting-edge graphics features

Microsoft's Work Graphs is the star of Agility SDK 1.613.0. It's a revolutionary new API that changes how GPUs are utilized. Work Graphs lets developers generate and schedule GPU workloads on-the-fly, with no reliance on the CPU host. This enables higher performance, lower latency, and greater scalability for tasks like:

• Culling

• Binning

• Chaining compute work

Dynamic work expansion, previously difficult to achieve, is now possible with Work Graphs.

New shader capabilities in this latest iteration include:

• Start Vertex/Instance Location - Allows efficient address calculations based on API draw call parameters.

• Wave Size Range - Developers can specify preferred wavefront sizes in shaders, reducing the need for multiple shader entry points.

• Expanded Comparison Sampling - Provides more options for evaluating nearby pixels, enabling better cross-platform shader creation.

This brings faster CPU-to-GPU data transfers by utilizing GPUs' resizable memory mapping (BAR) to bypass the PCI bus bottleneck. However, it requires Windows 11 Build 26080 or later.

Other Enhancements

• Incrementing Constants in ExecuteIndirect - Root signature constants increment per command

• Generic Programs in State Objects - Extends shader definitions across vertex, mesh, and compute pipelines

Read More Here

OpenAI releases Transformer Debugger

As language models grow larger and more capable, there is an increasing need for tools to interpret their behavior and understand their inner workings. OpenAI has released the Transformer Debugger to help meet this need.

The Transformer Debugger combines automated interpretability techniques with sparse autoencoders. It allows intervening in a model's computations and seeing impacts on behaviors like:

• Next-word prediction

• Attention patterns

It identifies key model components driving behaviors and traces connections between components.

What's Included:

• Neuron viewer React app to host the debugger

• Backend activation server

• Inference libraries for GPT-2 with access to model activations

• Datasets of top examples that highly activate different model components

While currently focused on small language models, tools like the Transformer Debugger may become essential as models grow in complexity and importance. It enables researchers to inspect, explain, and ultimately govern the behavior of large language models. The open-source availability allows broad exploration and extension by the AI community.

Read More Here

🔥 More Notes

• Context-switching - one of the worst productivity killers in the engineering industry

• Bebop v3: a fast, modern replacement for Protocol Buffers

• How Khan Academy Scaled to 30 Million Users

Youtube Spotlight

Jensen Huang, Founder and CEO of NVIDIA

Jensen Huang, the Founder and CEO of NVIDIA, discusses his journey from joining LSI Logic to founding NVIDIA. He talks about the motivation behind leaving LSI Logic to start a company with a new mission to solve problems that normal computers can’t. He also highlights the challenges faced by the company, such as financial crises and market competition, and the key role of early indicators of future success in driving the company forward.

Was this forwarded to you? Sign Up Here