Priors

TOP OF MIND

NVIDIA's Nemotron 3 Nano Omni model release unifies vision, audio, and language for up to 9x more efficient AI agents, signaling a significant leap in multimodal AI capabilities. China blocks Meta's acquisition of AI startup Manus, asserting origin-country jurisdiction and highlighting the complexities of global AI governance. OpenAI reportedly misses revenue targets, raising concerns about the AI bubble and the long-term viability of AI business models.

NVIDIA WATCH

NVIDIA launches Nemotron 3 Nano Omni, a multimodal AI model up to 9x more efficient than previous models. This release marks a significant milestone in NVIDIA's AI roadmap, unifying vision, audio, and language capabilities in a single model. The Nemotron 3 Nano Omni model is now available on Amazon SageMaker JumpStart, expanding its reach to a broader audience of developers and researchers.

MODEL RELEASES THAT MATTER

• NVIDIA Nemotron 3 Nano Omni: a multimodal AI model up to 9x more efficient than previous models, unifying vision, audio, and language capabilities.
• Mistral Medium: an upcoming model release from Mistral AI, promising improved performance and capabilities.

QUOTE OF THE DAY

Sundar Pichai, CEO of Google, marks 20 years of Google Translate, now powered by Gemini models for real-time conversations. "Hello. How are you? Thank you. I love you. Please." Pichai highlights the significance of Google Translate in bridging language gaps and facilitating global communication.

POLICY & SUPPLY CHAIN

China blocks Meta's acquisition of AI startup Manus, asserting origin-country jurisdiction and highlighting the complexities of global AI governance. The EU orders Google to open Android to third-party AI assistants, which Google opposes as unwarranted intervention. Experts warn of expanding powers to block online speech in India, citing concerns about transparency and oversight.

GPU PRICING PULSE

No pricing data available in today's signals.

DEV TOOLS & RESEARCH

Research presents a method for training large AI models using hyperparameter-divergent ensembles with automatic learning rate exploration. Introduces ZenBrain, a neuroscience-inspired 7-layer memory architecture for autonomous AI systems.

PULSE

• Kinetix AI unveils KAI, a humanoid robot with unprecedented degrees of freedom, hybrid dexterous hand, and 18,000 sensors for human-like interaction. (reddit-singularity)
• Google signs classified AI deal with Pentagon amid employee opposition. (reddit-singularity)
• Accenture deploys over 740,000 M365 Copilot seats, Microsoft's largest deployment to date. (x-satyanadella)

AI ECOSYSTEM

• [FT] The great American data centre divide: US data center regulations driving investment shifts to more lenient regions.
• [TechCrunch] OpenAI ends Microsoft legal peril over its $50B Amazon deal, resolving a significant partnership dispute.
• [TechCrunch] China blocks Meta's $2B Manus deal after months-long probe, highlighting the complexities of global AI governance.
• [Stratechery] An Interview with OpenAI CEO Sam Altman and AWS CEO Matt Garman About Bedrock Managed Agents, discussing the future of AI development and deployment.

WHAT TO WATCH TOMORROW

No specific calendar items mentioned in today's signals.

AI-synthesized. Experimental. Not investment advice. Not for circulation.