The Rundown: Google DeepMind just unveiled AlphaChip, an AI system that designs computer chips by using reinforcement learning to create superhuman chip layouts in hours rather than months.

The details:

AlphaChip has been used to design layouts for the last three generations of Google’s Tensor Processing Units (TPUs), improving performance and accelerating the design cycle.

The AI system uses a novel “edge-based” graph neural network to learn relationships between chip components and generalize across different chip designs.

Google is releasing a pre-trained checkpoint of AlphaChip, sharing the model weights to encourage further research and development in AI-assisted chip design.

AlphaChip’s impact extends beyond Google, with companies like MediaTek adopting the technology for their most advanced chips used in smartphones and other devices.

Why it matters: AlphaChip creates a powerful feedback loop: better AI models design better chips, which enable the training of even more advanced AI models, which design even better chips, and so on. This self-reinforcing cycle could dramatically accelerate AI progress, and at a level that any human will struggle to compete with.

Image source: Google

The Rundown: Google just upgraded its NotebookLM tool, adding support for YouTube videos and audio files, along with easier sharing of Audio Overviews—its latest viral AI hit that turns notes, PDFs, Google Docs, and more into AI-generated podcasts.

The details:

NotebookLM now supports public YouTube URLs and audio files, allowing users to analyze videos, lectures, and audio alongside existing text sources.

The tool leverages Gemini 1.5’s multimodal capabilities to summarize key concepts from videos and transcribe audio.

A new sharing feature allows users to generate public links for Audio Overviews, making collaboration even easier.

These updates aim to streamline tasks such as creating study guides, analyzing multiple perspectives on issues, and extracting important information from video, audio, and text.

Why it matters: It’s a big day for Google. The company’s viral hit with NotebookLM is now even more impressive with access to YouTube videos and audio files. YouTube is an endless treasure chest of how-to guides, lectures, documentaries, and entertainment—and now, anyone can consume hours worth of videos in minutes with AI.

Image source: University of Yamagata

The Rundown: Archaeologists from Japan’s Yamagata University, in collaboration with IBM Research, used AI to uncover 303 previously unknown geoglyphs near Peru’s famous Nazca Lines, nearly doubling the number of known figures at the site.

The details:

The newly discovered geoglyphs, dating back to 200 BC, depict various animals and humans, including parrots, cats, monkeys, killer whales, and even decapitated heads.

AI combined with low-flying drones dramatically accelerated the discovery process, accomplishing nearly a century’s worth of work in six months.

These smaller geoglyphs (10-25 feet across) provide new insights into the transition from the Paracas culture to the Nazca culture.

The findings, published in the Proceedings of the National Academy of Sciences, demonstrate AI’s ability to help greatly improve archaeological research.

Why it matters: Is there anything AI can’t help us accomplish? The amount of time saved using low-flying drones and artificial intelligence is worth repeating: 100 years worth of work in six months. The ways in which AI is going to impact our lives are still vast and largely unknown, as this discovery proves.