AMD claims 4x energy efficiency gain in AI systems over two years
AMD has reported significant progress in improving the energy efficiency of its AI systems. The company claims its latest systems are already four times more energy-efficient compared to models from two years ago. This brings AMD closer to its goal of making AI computing less power-hungry.
Meanwhile, Waymo has developed a custom chip for its self-driving cars to stay ahead of Tesla. The 5nm machine learning accelerator promises over 1,000 TOPS of performance and ultra-low latency, positioning it as a key differentiator in the autonomous vehicle market.
On the security front, Microsoft is facing a serious issue: its patches failed to fix an on-premises SharePoint vulnerability, which is now being exploited in zero-day attacks. The company has acknowledged the problem but has not yet provided a complete fix. Additionally, Russian hackers are posing as Signal support to launch phishing attacks, targeting users of the encrypted messaging app.
At the DEF CON conference, the Franklin project was announced, aiming to enlist hackers to help harden critical infrastructure. Reports on voting security will also be presented, highlighting ongoing concerns about election integrity.
In the software world, Go version 1.27 has been released, expanding support for generics to methods. This update may delight diehard Go developers but could displease AI overlords, as the new features are not optimized for machine learning workloads.
AI companies are facing accusations of burning books. Advocates have filed a complaint with the FTC, alleging that companies are destroying physical books to train their AI models. The practice, described as "Fahrenheit 203" — a reference to the temperature at which GPUs stop processing literature — has drawn criticism from authors and publishers.
AMD's efficiency gains are part of a broader trend in the industry to reduce the environmental impact of AI. The company's "House of Zen" architecture is credited with the improvements, though specific benchmarks were not disclosed. Waymo's chip development, meanwhile, underscores the growing competition in the autonomous vehicle space, where Tesla has been a dominant player. The new chip is expected to enhance Waymo's ability to process sensor data in real-time, improving safety and performance.
