AI: Weekly Summary. RTZ #666

Michael Parekh AI: Reset to Zero

4 weeks ago

4 MIN READ

Nvidia GTC 2025 rich in tech roadmap: Nvidia’s 2025 GTC keynote by founder/CEO Jensen Huang was a rich and deep roadmap on the company’s AI GPU direction over the next few years. From the current Hopper to the ramping Blackwell to Rubin and then Feynman, the company has a unique roadmap to AI GPUs and infrastructure at Scale for both AI training and inference in all its evolving forms. The company did a good job explaining how their approach addresses the ‘extreme AI computing’ problem (see chart below) of delivering ever rising amounts of ‘intelligence tokens’ efficiently for a unit of power, optimizing for throughput and response time for users. Nvidia also showed off its product roadmap for AI Robotics, self-driving cars, and local at the edge AI computers and devices, leveraging ‘Small AI’. Company continues to have a leading position despite rising competitive concerns in the market. More here.

Nvidia beefs up Synthetic Data capabilities: Nvidia also announced a $320 million acquisition of Gretel in the ‘Synthetic Data’ space, buttressing its software capabilities in a critical input area to scale LLM AIs over the next few years. Synthetic Data is a key item especially for the ‘Physical AI’ world of robotics and self-driving cars, as well as potentially solving the ‘Peak Data’ problem discussed by AI luminaries like ex OpenAI co-founder Ilya Sutskever and others. Both synthetic data and synthetic content, are going to be key drivers of AI scaling across reasoning and agents, in the next few years. Nvidia’s moves bolster their capabilities in this critical area going forward relative to their product roadmap. More here.

Nvidia’s ‘hundreds of billions’ in US Investments: Nvidia also outlined its plans to invest ‘hundreds of billions’ of dollars in the US with its partners TSMC, Foxconn and others, accelerating its contributions to ‘reshoring’ the AI supply chain away from China, with more investments closer to the US. This continues a similar trend by a range of big players including Apple, Softbank/OpenAI Stargate, TSMC and others, given the current administration’s laser focus on this issue via tariff policies. Separately, Microsoft also announced a multi billion dollar expansion of its AI Infrastructure and Power investments with Blackrock, the UAE’s MGX and other energy partners, with more AI Infrastructure and Power investments in the US. More here.

Apple AI Siri changes: Apple responded to its latest delays with the AI upgrades to Siri with management changes. The Siri responsibility was moved from AI boss John Giannandrea (JG), to a senior Apple executive Mike Rockwell who previously led the Vision Pro effort. Apple continues to focus on its efforts and delays around Apple Intelligence, and focusing on a bottom up AI applications and services efforts across its hardware and software platforms, optimizing them for user privacy and personalized applications for its billions of global users. The management moves begin to address the rising criticism and focus around Apple’s AI efforts with Siri at the lead. Expect more Siri and Apple Intelligence features later this year and next. Separately, Amazon is executing a similar AI effort with its Alexa/Echo services and devices. More here.

DC Decides AI Policies: The administration is closing in on finalizing its AI policies in the coming weeks, adjusting previous Biden era policies, and implementing its priorities on AI infrastructure and investments in the US going forward. Key upcoming decisions include the divisive issue of ‘Fair Use’ for copyrights in AI training, as well as potentially modifying the Biden era ‘AI Diffusion Rules’ in how US Ai and tech companies can sell their products and services into a globally tiered marketplace. Also up are US policies on streamlining AI Infrastructure and Power rollouts, balancing federal, state and local priorities through a thicket of cross-industry issues. More here.

More like this