March 2024 Summaries

In celebration of Women's History Month, the text highlights four inspiring women who have created impactful apps and games that address various societal issues. Múkami Kinoti Kimotho, from the USA, developed Royelles, a game aimed at inspiring girls and non-binary people to pursue careers in STEAM fields, using engaging characters and narratives. In Indonesia, Leonika Sari Njoto Boedioetomo founded Reblood, an app that connects blood donors with recipients, addressing the blood donation shortage and promoting healthier lifestyles. In Brazil, Luciane Antunes dos Santos and Renato Hélio Rauber created Carsul, an urban mobility app focused on safety and security, inspired by personal tragedy to prevent accidents and assist patients in need of transportation. Lastly, in France, Diariata N'Diaye launched App-Elles, a safety app for women that provides a platform for sending alerts and accessing support networks, with additional plans for a discreet wearable alert device. These initiatives are part of the #WeArePlay stories, showcasing innovation and empowerment driven by women.

The Build with Google AI video series returns for a second season, offering developers practical projects to leverage Google's AI technology, specifically the Gemini API, for advanced reasoning, content generation, and image interpretation. This season aims to help developers integrate AI into their workflows by providing customizable code projects that serve as templates for solving business problems. The series introduces new applications like an AI Slides Reviewer for Google Workspace, a Flutter Code Agent for generating user interface designs, and an AI Data Agent for translating questions into code. Developers can also explore upgraded tutorials from the first season, now enhanced with the Gemini API, and gain insights from the creators of these AI projects. Google's AI developer resources have a new home at ai.google.dev, offering a comprehensive platform for building AI-driven applications.

Google AI Studio has introduced Gemini Tuning, enabling developers to customize the Gemini 1.0 Pro model for specific needs using their own data. This advancement allows for higher-quality output than few-shot prompting, using Parameter Efficient Tuning (PET) to deliver customized models with reduced latency and without the extensive resources required for traditional fine-tuning. PET requires only a few hundred data points, easing the data collection burden, and allows tuning for tasks such as classification, information extraction, and structured output generation. The tuning process in Google AI Studio is straightforward, requiring no advanced engineering skills, and can be initiated with as few as 20 examples, though a dataset of 100 examples is recommended for optimal performance. Additionally, developers can utilize the Gemini API to perform tuning, offering flexibility in integrating custom models into various applications.

Google I/O is set for May 14th, inviting attendees to participate online in an event aimed at developers interested in application development and modernization. This year's focus is on the Gemini era, which presents new opportunities for creating AI-enabled applications. The event promises insights into accelerating the journey from ideas to production-level AI applications, with updates on mobile, web, and multiplatform development and strategies for cloud scalability. Participants can engage with over 100 sessions, workshops, codelabs, and demos to explore their areas of interest. To stay updated on Google I/O and related events, registration on the Google I/O site is recommended, and the livestreamed keynotes commence at 10 am PT on May 14th.

Gema Parreño Piqueras, an AI/ML Google Developer Expert from Madrid, Spain, is celebrated for her contributions to Artificial Intelligence and Machine Learning, particularly her innovative project Deep Asteroid, which assists NASA in tracking asteroids. Her unique blend of architectural skills and expertise in data science has led her to become a significant figure in the tech world, especially for women aspiring to enter the field. As a data scientist at Izertis and an active participant in the Google Developers community, Gema's journey highlights the importance of community support and the willingness to experiment and learn from failure. Her involvement in the GDE program has been instrumental in her personal and professional growth, allowing her to collaborate with Google teams and attend numerous events across Europe. Through her work, Gema emphasizes the power of persistence and the necessity of finding supportive networks like the GDE program for learning and development in technology.

Google for Games is set to participate in the Game Developers Conference (GDC) in San Francisco from March 18 to March 22, offering a comprehensive suite of tools and hosting eight sessions for game developers, designers, and business teams. These sessions will cover topics such as scaling game development using Firebase, Android, and Google Cloud tools, enhancing game performance with Vulkan and Android Adaptability Technology, and creating global-scale games with Google Cloud. Additionally, Google will focus on increasing user engagement through generative AI and Google Play's new features, maximizing reach across screens by expanding mobile games to PC using platforms like Flutter, and offering insights into Android input on desktop. The conference aims to provide attendees with knowledge on leveraging Google's technologies and services to improve game development, player experience, and business growth within the gaming industry.

Since its introduction in 2017, TensorFlow Lite has enabled on-device machine learning, with MediaPipe enhancing its capabilities in 2019 by supporting full ML pipelines. The experimental release of the MediaPipe LLM Inference API marks a significant advancement, allowing Large Language Models (LLMs) to function fully on-device, despite their demanding memory and compute requirements. This new API is optimized for web, Android, and iOS platforms, initially supporting models like Gemma, Phi 2, Falcon, and Stable LM. It benefits from various optimizations, including new ops, quantization, and weight sharing, to manage the computational intensity of LLMs. On Android, the API is currently intended for experimental use, with production applications advised to use the Gemini API or Gemini Nano through Android AICore, a feature in Android 14 that supports high-end devices with ML accelerators and safety filters. The API facilitates on-device LLM integration for developers, allowing them to prototype and test LLMs using platform-specific SDKs. Significant optimizations across MediaPipe, TensorFlow Lite, and XNNPack have been implemented to support efficient LLM inference, with strategies like sharing weights, optimizing fully connected operations, and employing custom GPU operations to balance compute and memory demands. The release aims to expand further throughout 2024, with plans for more platforms, models, and broader conversion tools, signaling a transformative step for on-device machine learning.