AI News: Gemini 3 Rumors, Apple M5, Meta Layoffs, Google's Quantum Breakthrough, and more!

Google has launched a new vibe coding product built on Gemini, allowing users to create AI applications from idea to production with minimal coding.

• Google's new vibe coding product leverages Gemini to build AI applications from scratch with minimal coding.

• Apple's M5 chip significantly boosts local AI inference across its updated product lineup.

• Google's quantum computer achieved verifiable quantum advantage, demonstrating a massive speedup for specific molecular computations.

Google unveiled a new AI-powered vibe coding product based on Gemini, enabling users to rapidly build web applications by simply describing their desired outcome, exemplified by a phone wallpaper generator created in 78 seconds. Apple refreshed three major products with its new M5 chip, significantly boosting local AI inference capabilities and signaling a stronger push into artificial intelligence. Meanwhile, Google's quantum computer achieved verifiable quantum advantage, demonstrating a breakthrough algorithm that vastly outperforms classical computing for specific molecular interactions.

Google's Vibe Coding Product

• 00:00:04 Google has launched a new 'vibe coding' product built on Gemini, allowing users to describe an application and have it built from idea to production within the browser. This tool is specifically designed for AI applications, enabling rapid development and deployment, as demonstrated by a mobile-first web app for generating phone wallpapers created in just 78 seconds. Future updates will incorporate database setup, file storage, authentication, and API building to enhance its capabilities.

Apple's M5 Chip & AI

• 00:05:46 Apple has refreshed the 14-inch MacBook Pro, iPad Pro, and Apple Vision Pro with its new M5 chip, which offers over four times the peak GPU compute performance for AI compared to the M4. This chip includes a next-generation GPU with a neural accelerator in each core, a more powerful CPU, a faster neural engine, and higher unified memory bandwidth. While competitive with NVIDIA DGX for pure inference, Apple still needs to heavily invest in developing its own robust artificial intelligence capabilities, like a better version of Siri, to fully leverage its powerful hardware.

Andre Karpathy's LLM Critique

• 00:08:08 Andre Karpathy's interview highlights significant limitations of current Large Language Models (LLMs), stating they lack sufficient intelligence, multimodality, computer interaction, and memory, describing them as 'cognitively lacking.' He argues LLMs mainly offer a 'hazy recollection of the Internet,' relying too much on memorization and struggling with generalization or creating truly novel ideas, comparing their current state to only recreating 'cortical tissue' without the rest of the brain's functions like memory or emotions. While some disagree, acknowledging ongoing solutions like 'scaffolding' and 'tool use,' the core model's inherent limitations remain a point of discussion.

Google's Quantum Breakthrough

• 00:11:00 Google's Willow quantum chip has achieved the first verifiable quantum advantage by running a new 'quantum echoes' algorithm 13,000 times faster than the best classical algorithm on a supercomputer. This breakthrough allows for explaining interactions between atoms in a molecule using nuclear magnetic resonance, with potential applications in drug discovery and material science. The verifiable nature of this result signifies a crucial step towards real-world applications of quantum computing, though some skepticism persists regarding the practical implications of such 'contrived results.'