Google I/O Recap 🇬, AI Detecting Marine Plastic Pollution 🌊, Benchmark on GPT-4o Breakthrough 🕶️

🤯 Game Changers

The most impactful articles of the day

Everything Google Announced at I/O 2024: Gemini, Search, Project Astra, Scam Detection

Google I/O 2024 was packed with AI announcements:

• Project Astra as their future of AI assistants

• VEO text-to-video

• Gemini Flash the lighter model for high-frequency tasks

• Gemini Pro has waitlist to preview a 2 million context window

• PaliGemma, new lightweight open vision-language model

• SynthId, AI content watermarking

• Gemini Nano on Chrome (early preview) and on mobile with scam detections

• Kotlin Multiplatform support

• Project IDX open beta, an AI-assisted workspace for full-stack development

• and much more.

🧐 Daily Picks

Curated picks and most shared articles on techtok.today

GPT-4o's Memory Breakthrough – Needle in a Needlestack

GPT-4o had a breakthrough score for Needle in a Needlestack— a new benchmark to measure how well LLMs pay attention to the information in their context window, signifying improvement in memory capabilities, and higher efficiency in storing and retrieving large amounts of information.

CERN's edge AI data analysis techniques used to detect marine plastic pollution

The EU project Edge SpAIce applies CERN’s edge AI technology to detect and track marine plastic pollution by acquiring and processing high-resolution pictures from satellites in near real-time, aiming to provide useful data for clean-up operations and improve our understanding of plastic pollution patterns.

Glider – open-source eInk monitor with an emphasis on low latency

Glider, an open-source e-ink monitor, emphasizes low latency and supports a wide range of e-ink screen sizes from 4.3 to 42 inches, with features like partial update, greyscale display, and dithering to improve image quality.

✨ "Too long don't README"

GitHub trending repos of today

BasedHardware/OpenGlass

OpenGlass is an open-source smart glasses project that can be built for less than $25 using off-the-shelf components, allowing users to record their life, remember people they meet, identify objects, and translate text.

pipecat-ai/pipecat

Pipecat is a framework for building voice and multimodal conversational agents, such as personal coaches, meeting assistants, story-telling toys, customer support bots, and more, with support for WebRTC, text-to-speech, and voice activity detection.

HazyResearch/ThunderKittens

ThunderKittens is a framework for writing fast deep learning kernels in CUDA, focusing on simplicity, extensibility, and speed, with features like tensor core support, shared memory management, and distributed shared memory.

📱 Product Picks

Curated products from Product Hunt

Wegic

Hi, I'm Wegic, the first AI web designer and developer by your side, powered by the latest GPT-4o model. I can help you create and modify websites in tons of languages through simple conversations. Feel free to chat with me and I can bring your ideas to life!

fynk

fynk is more than just contract management. Built with powerful AI at its core, it seamlessly imports and integrates your existing and external contracts into powerful workflows: create, automate, manage, collaborate, and sign - all in one platform!

SaveDay

Instantly summarize and ask questions about saved articles and YouTube videos on your phone. Then save them to your knowledge hub to easily find them later.

Today from Engineering Blogs

Articles from engineering blogs of big tech companies

From Predictive to Generative – How Michelangelo Accelerates Uber’s AI Journey

Uber's centralized ML platform Michelangelo has been instrumental in driving its ML evolution, enabling the company to transition from using simple predictive models to advanced deep learning and generative AI models across its business lines, with over 400 active ML projects, 20K monthly model training jobs, and 5K models in production serving 10 million real-time predictions per second.