Truncated: latest AI news

Thursday 5 November

F3 Lightning: How Meta Accelerates Ads ML Features to Production at Scale

AI-generated summary Engineering at Meta explains that F3 Lightning was developed to allow every feature to be released on its own schedule while preserving extraction efficiency. Engineering at Meta notes that the system compiles each feature independently and consolidates shared work at startup using identifiers based on the computations themselves.

The AI boom is turning fiber-optic cables into hot commodities

AI-generated summary Corning is investing $585 million in expanding manufacturing in North Carolina to meet increasing demands for fiber-optic cables from companies such as Amazon, Meta, and Nvidia. Relativity Networks developed a hollow-core fiber called ChronoCore and partnered with Prysmian to scale production and connect data centers faster.

Also covered by forbes.com

"Used 5% of the time": AI already made your RAM pricier, and now OpenAI's Sam Altman says it's wasted in your Windows PC

AI-generated summary OpenAI CEO Sam Altman stated on a podcast that keeping limited DRAM in personal computers used five percent of the time is quite terrible for global allocation. Microsoft executive Satya Nadella instead promotes unmetered intelligence with models running locally on Windows PCs.

Also covered by theverge.com XDA videocardz.com
+6 moreEngadget tomshardware.com Decrypt AIStockWire techradar.com CNET

Data Points: AI models can read other models’ weights

AI-generated summary MIT researchers report that OpenAI model GPT-6 Astra can read the raw weights of other AI models and rewrite them without gradient-based training. Anthropic and academic collaborators released FlashProve, an open-source system that produces cryptographic proofs that a response came from a specific committed model.

Earlier: Progress on provable inference for large language models