Welcome to Get The Gist, where every weekday we share an easy-to-read summary of the latest and greatest developments in AI—news, innovations, and trends—all delivered in under 5 minutes! ⏱

In today’s edition:

• NVIDIA Unveils Nemotron AI Models

• Elon Musk Announces Grok 3 LLM

• Instagram Tests Creepy AI-Generated Images of Users

• Meta Open-Sources its Byte Latent Transformer (BLT)

• And more AI news….

Top Developments

Nvidia

NVIDIA Unveils Nemotron AI Models

The Gist: NVIDIA has introduced the Llama Nemotron and Cosmos Nemotron model families, designed to supercharge agentic AI applications by enabling efficient and scalable AI agents for diverse industries.

Key Details:

• Llama Nemotron models are optimized for agentic AI, excelling in tasks like chat, coding, and decision-making, with scalable options (Nano, Super, Ultra).

• Cosmos Nemotron models integrate vision and language processing, enabling AI agents to analyze images and video for applications in robotics, healthcare, retail, and more.

• Models are available as microservices via NVIDIA AI Enterprise, supporting deployment across PCs, data centers, and the cloud.

• Enterprises can customize and integrate these models with their own data using NVIDIA NeMo tools for streamlined development.

Grok

Elon Musk Announces Grok 3 LLM

The Gist: xAI's Grok 3 large language model has completed pretraining, utilizing 10 times the compute resources of its predecessor on a 100,000 Nvidia H100 GPU supercluster.

Key Details:

• Grok 3 was pretrained on xAI’s Colossus supercomputer, optimized for large-scale AI with 100,000 Nvidia H100 GPUs, soon expanding to 200,000 GPUs.

• Training leveraged user-generated data from the X platform, ensuring a more powerful and sophisticated language model.

• Colossus will eventually host over a million GPUs to train even larger models with trillions of parameters, pushing closer to artificial general intelligence.

• xAI aims to stay competitive with OpenAI, Google DeepMind, and Anthropic by accelerating model training and deploying cutting-edge AI systems.

Instagram

Instagram Tests Creepy AI-Generated Images of Users in Feeds

The Gist: Instagram is testing a feature where AI-generated portraits of users appear in their feeds after using Meta AI's "Imagine Yourself" tool, sparking user discomfort and skepticism.

Key Details:

• Meta AI’s "Imagine Yourself" feature creates AI-generated images of users unprompted, visible only to the individual user, after they engage with the tool.

• Instagram labels these images as "only visible to you," and Meta confirms this is an intentional part of a limited test.

• The feature follows Meta’s recent push into AI-generated content but raises concerns about user consent and transparency.

• This marks the first instance of a social platform inserting AI-generated user images into feeds without explicit instructions, testing boundaries of AI integration.

Quick Gist

LG Electronics has partnered with Microsoft to develop AI agents for use in homes, vehicles, hotels, and offices, integrating Microsoft’s speech technologies into LG's AI Home Hub project. (Read More)

Meta announced the appointment of Dana White, a key ally of Donald Trump, to its board of directors. Meta believes White will play a critical role in shaping its AI initiatives. (Read More)

Google is beta testing plans to integrate its Gemini AI assistant into Wear OS, enhancing smartwatch interactivity with a more conversational experience. (Read More)

Google DeepMind has established a new AI research team, led by Tim Brooks, to develop "world models" that simulate real-world environments, targeting advancements in artificial general intelligence. (Read More)

Alphabet plans to integrate Gemini AI into Google TV's voice control system, enhancing natural command recognition and improving content search. The rollout is set for late 2025. (Read More)

Google is also assembling a new team led by Tim Brooks to develop AI models capable of accurately simulating the physical world, aimed at creating tools for real-time interactive generation. (Read More)

Apple plans to update its iPhone notification system to indicate when AI-generated summaries are used, addressing concerns about inaccuracies in its current implementation. (Read More)

Fortune Media has trademarked the term FORTUNE, emphasizing its established presence and rights both in the U.S. and globally. (Read More)

Meta has open-sourced the Byte Latent Transformer (BLT), a new large language model architecture that processes raw byte data efficiently, achieving comparable performance to Llama 3 with 50% fewer inference FLOPS. (Read More)

THAT’S IT FOR TODAY!