April 2024 Summaries

New AI-powered tools in open beta are being introduced to enhance the Google for Developers documentation and learning resources, aiming to streamline searches and simplify coding processes for developers. Users can interact with an AI-powered chat to ask questions and access documentation, use Code Explain for natural language explanations of code snippets, and explore APIs through the API Explorer without needing to code. Additionally, AI-powered search provides quick, relevant responses by pulling text and links from across the site, facilitating faster answers to developers' queries. Access to these features requires a Google Developer Profile, and the tools are currently available on desktop under Google's Generative AI Additional Terms of Service.

Continuous Integration and Continuous Deployment (CI/CD) are fundamental practices in modern software development that enhance efficiency, quality, and speed by automating code integration, testing, and deployment processes. They are designed to detect issues early, thus reducing costs, improving reliability, and enabling faster release cycles by shifting issue detection left in the development lifecycle. The introduction of Checks privacy compliance CI/CD tooling further advances this automation by integrating privacy and compliance checks into the CI/CD pipeline, allowing for automated scanning of app compliance against global regulations and company policies. By incorporating Checks, teams can receive real-time alerts on compliance issues, understand data sharing practices, and ensure that new builds adhere to privacy policies, which not only maintains high compliance standards but also allows for quick and confident deployments. This proactive approach is facilitated through seamless integration with platforms like GitHub, Jenkins, and others, helping development and compliance teams align with regulatory requirements and safeguard user data.

Startups are increasingly leveraging AI to tackle significant global challenges, and to support this innovation, Google has launched the Google for Startups Accelerator: AI-First program in North America. This inaugural initiative is designed for U.S. and Canadian startups that integrate AI into their core services, offering them 10 weeks of mentorship, technical support, and access to Google's latest AI tools, including Gemini. The program encompasses workshops on diverse topics such as tech infrastructure, UX, product growth, sales, leadership, and OKRs. The first cohort consists of 15 AI startups focused on various fields, from healthcare and cybersecurity to legal practice management and fintech. Following a kickoff at Cloud Next 2024 and culminating in a Demo Day in June, the startups will join a broader Google for Startups community for ongoing support and networking. The program aims to foster responsible scaling of AI solutions while building a network of innovative founders, mentors, and investors.

Gemma, a family of lightweight open models announced in February, has expanded with two new variants, CodeGemma and RecurrentGemma, designed to enhance machine learning capabilities in code completion and research experimentation. CodeGemma provides developers with advanced code completion, generation, and chat functions, available in multiple pretrained and instruction-tuned variants, supporting languages like Python and JavaScript, and optimized for both local and cloud environments. RecurrentGemma, distinct from traditional Transformer models, utilizes recurrent neural networks for efficient memory usage and faster inference at higher batch sizes, making it suitable for devices with limited memory. Both models are built on the principles of open availability and responsible AI, compatible with various software and hardware platforms including JAX, PyTorch, and NVIDIA GPUs. Alongside these models, Gemma 1.1 has been released with performance improvements and updated terms based on user feedback. The models are accessible on platforms like Kaggle, Hugging Face, and Vertex AI Model Garden, inviting developers to explore and innovate with these advanced tools.

Google AI has launched the Gemini 1.5 Pro model in public preview across 180+ countries via the Gemini API, introducing new features such as native audio understanding and a File API for easier file handling. Developers can now utilize system instructions and JSON mode for more controlled model output, and benefit from improvements in function calling. The model's input modalities have been expanded to include both audio and visual reasoning capabilities for video content. Additionally, a new text embedding model, "text-embedding-004," is available, offering superior retrieval performance compared to existing models. Developers are encouraged to access these advancements through Google AI Studio, explore the new Gemini API Cookbook for guidance, and participate in the community on Discord.

The ML Olympiad is a series of Kaggle Community Competitions designed to provide developers with practical machine learning experience by tackling real-world challenges, supported by ML GDE, TFUG, and other ML communities, with sponsorship from Google for Developers. In its third round, following successful iterations in 2022 and 2023, the Olympiad features over 20 active competitions, such as smoking detection through bio-signals, classification models for distinguishing jellyfish from plastic pollution, and predictive models for urban traffic density and weather conditions. Participants are encouraged to engage in these competitions to enhance their machine learning skills and contribute to diverse projects like CO2 emissions prediction, disaster response improvement, and machine translation, with each competition hosted by different experts and organizations across the globe. The initiative has previously attracted 605 teams, who participated in 32 competitions, generating numerous discussions and notebooks, and continues to invite new participants to explore a variety of challenges by searching "ML Olympiad" on Kaggle or following #MLOlympiad on social media.