google·Mar 19, 2025Original article

Loss of Pulse Detection on the Google Pixel Watch 3 (opens in new tab)

Google Research has developed a "Loss of Pulse Detection" feature for the Pixel Watch 3 to address the high mortality rates associated with unwitnessed out-of-hospital cardiac arrests (OHCA). By utilizing a multimodal algorithm that combines photoplethysmography (PPG) and accelerometer data, the device can automatically identify the transition to a pulseless state and contact emergency services. This innovation aims to transform unwitnessed medical emergencies into functionally witnessed ones, potentially increasing survival rates by ensuring timely intervention. ### The Impact of Witness Status on Survival * Unwitnessed cardiac arrests currently face a major public health challenge, with survival rates as low as 4% compared to 20% for witnessed events. * The "Chain of Survival" traditionally relies on human bystanders to activate emergency responses, leaving those alone at a significant disadvantage. * Every minute without resuscitation decreases the chance of survival by 7–10%, making rapid detection the most critical factor in prognosis. * Converting an unwitnessed event into a "functionally witnessed" one via a wearable device could equate to a number needed to treat (NNT) of only six people to save one life. ### Multimodal Detection and the Three-Gate Process * The system uses PPG sensors to measure blood pulsatility by detecting photons backscattered by tissue at green and infrared wavelengths. * To prevent false positives and errant emergency calls, the algorithm must pass three sequential "gates" before making a classification. * **Gate 1:** Detects a sudden, significant drop in the alternating current (AC) component of the green PPG signal, which suggests a transition from a pulsatile to a pulseless state, paired with physical stillness. * **Gate 2:** Employs a machine learning algorithm trained on diverse user data to quantify the probability of a true pulseless transition. * **Gate 3:** Conducts additional sensor checks using various LED and photodiode geometries, wavelengths, and gain settings to confirm the absence of even a weak pulse. ### On-Device Processing and User Verification * All data processing occurs entirely on the watch to maintain user privacy, consistent with Google’s established health data policies. * If the algorithm detects a loss of pulse, it initiates two check-in prompts involving haptic, visual, and audio notifications to assess user responsiveness. * The process can be de-escalated immediately if the user moves their arm purposefully, ensuring that emergency services are only contacted during true incapacitation. * When a user remains unresponsive, the watch automatically contacts emergency services to provide the individual's current location and medical situation. By providing a passive, opportunistic monitoring system on a mass-market wearable, this technology offers a critical safety net for individuals at risk of unwitnessed cardiac events. For the broader population, the Pixel Watch 3 serves as a life-saving tool that bridges the gap between a sudden medical emergency and the arrival of professional responders.

Machine LearningAIWearable TechnologyPhotoplethysmography+5

google·Mar 17, 2025Original article

Generating synthetic data with differentially private LLM inference (opens in new tab)

Researchers at Google have developed an inference-only method for generating differentially private (DP) synthetic data that avoids the high costs and data requirements associated with private fine-tuning. By prompting off-the-shelf large language models (LLMs) with sensitive examples in parallel and aggregating their outputs, the approach can generate thousands of high-quality synthetic data points while maintaining rigorous privacy guarantees. This method allows synthetic data to serve as a secure interface for model development, enabling teams to collaborate without requiring specialized knowledge of differential privacy. ## Differentially Private Prediction and Aggregation The core of this method relies on "private prediction," where privacy is applied to the model's output rather than the model itself. * Sensitive data points are distributed across multiple independent prompts, ensuring that no single individual's record can significantly influence the final output. * The LLM generates next-token predictions for each prompt in parallel, which are then aggregated to mask individual contributions. * The researchers designed a DP token sampling algorithm that treats the standard LLM "softmax" sampling process as a version of the exponential mechanism, a mathematical framework used to select the best option from a set while maintaining privacy. ## Enhancing Efficiency via KV Caching Previous attempts at private prediction were computationally expensive because they required a fresh batch of sensitive examples for every single token generated. * A new privacy analysis allows the system to reuse a fixed batch of sensitive examples across an entire generation sequence. * By maintaining the same context for each generation step, the system becomes compatible with standard inference optimization techniques like KV (Key-Value) caching. * This improvement enables the generation of synthetic data at a scale two to three orders of magnitude larger than prior methods. ## Optimizing Privacy Spend with Public Drafters To preserve the "privacy budget"—the limited amount of information that can be released before privacy is compromised—the method introduces a public drafter model. * The drafter model predicts the next token based solely on previously generated synthetic text, without ever seeing the sensitive data. * Using the sparse vector technique, the system only consumes the privacy budget when the public drafter’s suggestion disagrees with the private aggregate of the sensitive data. * This is particularly useful for structured data, where the drafter can handle formatting and syntax tokens, saving the privacy budget for the actual content. By leveraging off-the-shelf models like Gemma, this approach provides a scalable way to transform sensitive datasets into useful synthetic versions. These synthetic datasets are high-quality enough to replace real data in downstream machine learning tasks, such as in-context learning or fine-tuning models like BERT, without the risk of leaking individual user information.

Machine LearningAILarge Language ModelsDifferential Privacy+5

microsoft·Mar 3, 2025·2 min readCurated summary

How Microsoft Engineers Build AI: Learn about scalable RAG-enabled AI Apps

Microsoft’s new *How Microsoft Engineers Build AI* video series explains how its teams develop AI applications at scale. The first episode focuses on retrieval-augmented generation (RAG), using Copilot for Azure’s Ask Learn plugin as a practical example. It shows how RAG can combine proprietary data with large language models to deliver accurate, contextually relevant answers. ## Building AI Applications with RAG - RAG is presented as a practical way to improve AI applications without relying solely on model fine-tuning. - It retrieves relevant information from a knowledge base and provides that context to an LLM when generating responses. - The approach is useful for applications that need current, domain-specific, or proprietary information. ## The Ask Learn Plugin - Microsoft engineers explain how they built the Ask Learn RAG plugin for Copilot for Azure. - The plugin helps Azure developers find answers quickly within their existing workflow. - The project involved product managers and engineering leaders sharing development challenges, design decisions, and best practices. ## Challenges in Developing Reliable RAG - Selecting the right source content is essential for producing useful answers. - Data must be preprocessed effectively before it can be retrieved. - RAG systems require careful performance evaluation to measure accuracy and relevance. - Keeping responses accurate and up to date requires ongoing improvements to content and retrieval methods. ## Broader Microsoft Applications - The episode discusses RAG implementations across: - Copilot in Azure - Microsoft Security Copilot - Dynamics 365 Business Central - These examples demonstrate how RAG can support different products and business scenarios. The episode is intended as a practical introduction for developers building RAG-based applications, covering prototyping, data management, evaluation, and common pitfalls. Developers can explore the series alongside Microsoft Learn resources and Azure AI development tools such as Visual Studio and GitHub Copilot.

Read original(opens in new tab)

Machine LearningAILarge Language ModelsGithub Copilot+4

coupang·Nov 15, 2024Original article

Optimizing Logistics Receiving Processes Using Machine (opens in new tab)

Coupang has implemented a machine learning-based prediction system to optimize its logistics inbound process by accurately forecasting the number of trucks required for product deliveries. By analyzing historical logistics data and vendor characteristics, the system minimizes resource waste at fulfillment center docks and prevents operational delays caused by slot shortages. This data-driven approach ensures that limited dock slots are allocated efficiently, improving overall supply chain speed and reliability. ### Challenges in Inbound Logistics * Fulfillment centers operate with a fixed number of "docks" for unloading and specific time "slots" assigned to each truck. * Inaccurate predictions create a resource dilemma: under-estimating slots causes unloading delays and backlogs, while over-estimating leads to idle docks and wasted capacity. * The goal was to move beyond manual estimation to an automated system that balances vendor requirements with actual facility throughput. ### Feature Engineering and Data Collection * The team performed Exploratory Data Analysis (EDA) on approximately 800,000 instances of inbound data collected over two years. * In-depth interviews with domain experts and logistics managers were conducted to identify hidden patterns and qualitative factors that influence truck requirements. * Final feature sets were refined through feature engineering, focusing on vendor-specific behaviors and the physical characteristics of the products being delivered. ### LightGBM Implementation and Optimization * The LightGBM algorithm was selected due to its high performance with large datasets and its efficiency in handling categorical features. * The model utilizes a leaf-wise tree growth strategy, which allows for faster training speeds and lower loss compared to traditional level-wise growth algorithms. * Hyperparameters were optimized using Bayesian Optimization, a method that finds the most effective model configurations more efficiently than traditional grid search methods. * The trained model is integrated directly into the booking system, providing real-time truck quantity recommendations to vendors during the application process. ### Operational Trade-offs and Results * The system must navigate the trade-off between under-prediction (which risks logistical bottlenecks) and over-prediction (which risks resource waste). * By automating the prediction of necessary slots, Coupang has reduced the manual workload for vendors and improved the accuracy of fulfillment center scheduling. * This optimization allows for more products to be processed in a shorter time frame, directly contributing to faster delivery times for the end customer. By replacing manual estimates with a LightGBM-based predictive model, Coupang has successfully synchronized vendor deliveries with fulfillment center capacity. This technical shift not only maximizes dock utilization but also builds a more resilient and scalable inbound supply chain.

Machine LearningAIPredictive ModelingFeature Engineering+3

coupang·Nov 15, 2024Original article

Accelerating ML development through Cou (opens in new tab)

Coupang’s internal Machine Learning (ML) platform serves as a standardized ecosystem designed to accelerate the transition from experimental research to stable production services. By centralizing core functions like automated pipelines, feature engineering, and scalable inference, the platform addresses the operational complexities of managing ML at an enterprise scale. This infrastructure allows engineers to focus on model innovation rather than manual resource management, ultimately driving efficiency across Coupang’s diverse service offerings. ### Addressing Scalability and Development Bottlenecks * The platform aims to drastically reduce "Time to Market" by providing "ready-to-use" services that eliminate the need for engineers to build custom infrastructure for every model. * Integrating Continuous Integration and Continuous Deployment (CI/CD) into the ML lifecycle ensures that updates to data, code, and models are handled with the same rigor as traditional software engineering. * By optimizing ML computing resources, the platform allows for the efficient scaling of training and inference workloads, preventing infrastructure costs from spiraling as the number of models grows. ### Core Services of the ML Platform * **Notebooks and Pipelines:** Integrated Jupyter environments allow for ad-hoc exploration, while workflow orchestration tools enable the construction of reproducible ML pipelines. * **Feature Engineering:** A dedicated feature store facilitates the reuse of data components and ensures consistency between the features used during model training and those used in real-time inference. * **Scalable Training and Inference:** The platform provides dedicated clusters for high-performance model training and robust hosting services for real-time and batch model predictions. * **Monitoring and Observability:** Automated tools track model performance and data drift in production, alerting engineers when a model’s accuracy begins to degrade due to changing real-world data. ### Real-World Success in Search and Pricing * **Search Query Understanding:** The platform enabled the training of Ko-BERT (Korean Bidirectional Encoder Representations from Transformers), significantly improving the accuracy of search results by better understanding customer intent. * **Real-time Dynamic Pricing:** Using the platform’s low-latency inference services, Coupang can predict and adjust product prices in real-time based on fluctuating market conditions and inventory levels. To maintain a competitive edge in e-commerce, organizations should transition away from fragmented, ad-hoc ML workflows toward a unified platform that treats ML as a first-class citizen of the software development lifecycle. Investing in such a platform not only speeds up deployment but also ensures the long-term reliability and observability of production models.

Machine LearningAICI/CDMlops+5

coupang·Nov 14, 2024Original article

Optimizing the inbound process with a machine learning model (opens in new tab)

Coupang optimized its fulfillment center inbound process by implementing a machine learning model to predict the exact number of delivery trucks and dock slots required for vendor shipments. By moving away from manual estimates, the system minimizes resource waste from over-allocation while preventing processing delays caused by under-prediction. This automated approach ensures that the limited capacity of fulfillment center docks is utilized with maximum efficiency. ### The Challenges of Dock Slot Allocation * Fulfillment centers operate with a fixed number of hourly "slots," representing the time and space a single truck occupies at a dock to unload goods. * Inaccurate slot forecasting creates a binary risk: under-prediction leads to logistical bottlenecks and delivery delays, while over-prediction results in idle docks and wasted operational overhead. * The diversity of vendor behaviors and product types makes manual estimation of truck requirements highly inconsistent across the supply chain. ### Predictive Modeling and Feature Engineering * Coupang utilized years of historical logistics data to extract features influencing truck counts, including product dimensions, categories, and vendor-specific shipment patterns. * The system employs the LightGBM algorithm, a gradient-boosting framework selected for its high performance and ability to handle large-scale tabular logistics data. * Hyperparameter tuning is managed via Bayesian optimization, which efficiently searches the parameter space to minimize prediction error. * The model accounts for the inherent trade-off between under-prediction and over-prediction, prioritizing a balance that maintains high throughput without straining labor resources. ### System Integration and Real-time Processing * The trained ML model is integrated directly into the inbound reservation system, providing vendors with an immediate prediction of required slots during the request process. * By automating the truck-count calculation, the system removes the burden of estimation from vendors and ensures consistency across different fulfillment centers. * This integration allows Coupang to dynamically adjust its dock capacity planning based on real-time data rather than static, historical averages. To maximize logistics efficiency, organizations should leverage granular product data and historical vendor behavior to automate capacity planning. Integrating predictive models directly into the reservation workflow ensures that data-driven insights are applied at the point of action, reducing human error and resource waste.

Machine LearningAIPredictive ModelingBayesian Optimization+3

coupang·Nov 14, 2024Original article

Accelerating Coupang’s AI Journey with LLMs (opens in new tab)

Coupang is strategically evolving its machine learning infrastructure to integrate Large Language Models (LLMs) and foundation models across its e-commerce ecosystem. By transitioning from task-specific deep learning models to multi-modal transformers, the company aims to enhance customer experiences in search, recommendations, and logistics. This shift necessitates a robust ML platform capable of handling the massive compute, networking, and latency demands inherent in generative AI. ### Core Machine Learning Domains Coupang’s existing ML ecosystem is built upon three primary pillars that drive business logic: * **Recommendation Systems:** These models leverage vast datasets of user interactions—including clicks, purchases, and relevance judgments—to power home feeds, search results, and advertising. * **Content Understanding:** Utilizing deep learning to process product catalogs, user reviews, and merchant data to create unified representations of customers and products. * **Forecasting Models:** Predictive algorithms manage over 100 fulfillment centers, optimizing pricing and logistics for millions of products through a mix of statistical methods and deep learning. ### Enhancing Multimodal and Language Understanding The adoption of Foundation Models (FM) has unified previously fragmented ML tasks, particularly in multilingual environments: * **Joint Modeling:** Instead of separate embeddings, vision and language transformer models jointly model product images and metadata (titles/descriptions) to improve ad retrieval and similarity searches. * **Cross-Border Localization:** LLMs facilitate the translation of product titles from Korean to Mandarin and improve the quality of shopping feeds for global sellers. * **Weak Label Generation:** To overcome the high cost of human labeling in multiple languages, Coupang uses LLMs to generate high-quality "weak labels" for training downstream models, addressing label scarcity in under-resourced segments. ### Infrastructure for Large-Scale Training Scaling LLM training requires a shift in hardware architecture and distributed computing strategies: * **High-Performance Clusters:** The platform utilizes H100 and A100 GPU clusters interconnected with high-speed InfiniBand or RoCE (RDMA over Converged Ethernet) networking to minimize communication bottlenecks. * **Distributed Frameworks:** To fit massive models into GPU memory, Coupang employs various parallelism techniques, including Fully Sharded Data Parallelism (FSDP), Tensor Parallelism (TP), and Pipeline Parallelism (PP). * **Efficient Categorization:** Traditional architectures that required a separate model for every product category are being replaced by a single, massive multi-modal transformer capable of handling categorization and attribute extraction across the entire catalog. ### Optimizing LLM Serving and Inference The transition to real-time generative AI features requires significant optimizations to manage the high computational cost of inference: * **Quantization Strategies:** To reduce memory footprint and increase throughput, models are compressed using FP8, INT8, or INT4 precision without significant loss in accuracy. * **Advanced Serving Techniques:** The platform implements Key-Value (KV) caching to avoid redundant computations during text generation and utilizes continuous batching (via engines like vLLM or TGI) to maximize GPU utilization. * **Lifecycle Management:** A unified platform vision ensures that the entire end-to-end lifecycle—from data preparation and fine-tuning to deployment—is streamlined for ML engineers. To stay competitive, Coupang is moving toward an integrated AI lifecycle where foundation models serve as the backbone for both content generation and predictive analytics. This infrastructure-first approach allows for the rapid deployment of generative features while maintaining the resource efficiency required for massive e-commerce scales.

Machine LearningAILarge Language ModelsGenerative AI+5

coupang·Nov 14, 2024Original article

Meet Coupang’s Machine Learning Platform (opens in new tab)

Coupang’s internal Machine Learning Platform (MLP) is a comprehensive "batteries-included" ecosystem designed to streamline the end-to-end lifecycle of ML development across its diverse business units, including e-commerce, logistics, and streaming. By providing standardized tools for feature engineering, pipeline authoring, and model serving, the platform significantly reduces the time-to-production while enabling scalable, efficient compute management. Ultimately, this infrastructure allows Coupang to leverage advanced models like Ko-BERT for search and real-time forecasting to enhance the customer experience at scale. **Motivation for a Centralized Platform** * **Reduced Time to Production:** The platform aims to accelerate the transition from ad-hoc exploration to production-ready services by eliminating repetitive infrastructure setup. * **CI/CD Integration:** By incorporating continuous integration and delivery into ML workflows, the platform ensures that experiments are reproducible and deployments are reliable. * **Compute Efficiency:** Managed clusters allow for the optimization of expensive hardware resources, such as GPUs, across multiple teams and diverse workloads like NLP and Computer Vision. **Notebooks and Pipeline Authoring** * **Managed Jupyter Notebooks:** Provides data scientists with a standardized environment for initial data exploration and prototyping. * **Pipeline SDK:** Developers can use a dedicated SDK to define complex ML workflows as code, facilitating the transition from research to automated pipelines. * **Framework Agnostic:** The platform supports a wide range of ML frameworks and programming languages to accommodate different model architectures. **Feature Engineering and Data Management** * **Centralized Feature Store:** Enables teams to share and reuse features, reducing redundant data processing and ensuring consistency across the organization. * **Consistent Data Pipelines:** Bridges the gap between offline training and online real-time inference by providing a unified interface for data transformations. * **Large-scale Preparation:** Streamlines the creation of training datasets from Coupang’s massive logs, including product catalogs and user behavior data. **Training and Inference Services** * **Scalable Model Training:** Handles distributed training jobs and resource orchestration, allowing for the development of high-parameter models. * **Robust Model Inference:** Supports low-latency model serving for real-time applications such as ad ranking, video recommendations in Coupang Play, and pricing. * **Dedicated Infrastructure:** Training and inference clusters abstract the underlying hardware complexity, allowing engineers to focus on model logic rather than server maintenance. **Monitoring and Observability** * **Performance Tracking:** Integrated tools monitor model health and performance metrics in live production environments. * **Drift Detection:** Provides visibility into data and model drift, ensuring that models remain accurate as consumer behavior and market conditions change. For organizations looking to scale their AI capabilities, investing in an integrated platform that bridges the gap between experimentation and production is essential. By standardizing the "plumbing" of machine learning—such as feature stores and automated pipelines—companies can drastically increase the velocity of their data science teams and ensure the long-term reliability of their production models.

Machine LearningAICI/CDNatural Language Processing+5

figma·Jun 25, 2024·3 min readCurated summary

Ovetta Sampson on Inputs and Outputs | Figma Blog

Minimum viable data is the idea that AI projects should begin with representative, high-quality data—not with the most powerful model or feature. Ovetta Sampson argues that model outputs are overwhelmingly determined by their inputs, which reflect human choices and historical biases. Product builders should therefore question whether AI is necessary, who it serves, and whether the data is equitable enough to avoid harming overlooked groups. ## Data Quality Shapes AI Outcomes - The quality of an AI system depends primarily on the data used to train and operate it. - Decisions about what data to collect, exclude, label, and measure determine who the system recognizes and how it behaves. - Data is never purely objective: it is generated, engineered, and transformed by people. - Treating data as disconnected from human lives can produce “traumatized data sets,” embedding social, cultural, and economic harms into models. ## The Consequences of Omission - Historical datasets often exclude entire groups: - U.S. credit and mortgage models were developed before women could independently obtain mortgages or credit cards. - The U.S. Census did not recognize LGBTQ individuals until 2021, despite those people existing in earlier populations. - When people are absent from the data, models may fail to serve them or may expose them to harmful decisions. - The central question is not simply whether data exists, but whether it represents the people affected by the system. ## Define the Problem Before Choosing AI - Teams should first identify the problem they are trying to solve and determine whether ML or AI is appropriate. - The fact that a problem can be addressed with AI does not mean it should be. - Builders should ask: - Who is the product for? - Is the data equitable and sufficiently high quality? - What is the minimum data and technology needed? - Could the proposed solution increase human risks or reduce people to data points? - Minimum viable data means collecting what is necessary for a useful, responsible solution rather than indiscriminately gathering more data. ## Putting People Back in Control - Product builders and the public need to participate in decisions about how AI systems are designed and governed. - Important questions include who defines “good” data, who decides what enters a training set, and how much data is truly required. - Sampson recommends learning from work such as *Weapons of Math Destruction*, *Ghost Work*, and research on the lack of attention given to data work in AI development. AI development should start with the people affected by a system and the data needed to represent them fairly. Choosing the smallest appropriate dataset and validating its quality can be more responsible—and more effective—than pursuing larger models or unnecessary AI features.

Read original(opens in new tab)

Machine LearningAIGenerative AIData Quality+2

figma·Jun 17, 2024·3 min readCurated summary

Navigating the Promise and Pitfalls of AI | Figma blog | Figma Blog

AI’s promise is real, but useful AI products will emerge through experimentation rather than a race to ship features. Figma’s research suggests AI is already transforming individual workflows—especially for developers—while having a smaller effect on collaboration and foundational design work. To deliver lasting value, teams must focus on how AI reshapes products, industries, and group work, not just on model capabilities. ## Research and Methodology - Figma surveyed more than 1,800 designers, executives, and developers. - Participants came from the US, Canada, Australia, the UK, Japan, France, and Germany. - The survey ran from February 26 to March 3, 2024. - The report combines survey findings with discussions involving AI and design experts. - Its central premise is that AI’s impact depends heavily on product design and user experience—not only on the power of large language models. ## AI’s Uneven Transformation of Workflows - Developers were 60% more likely than designers to say AI had transformed the products they work on. - Developers use AI for daily tasks such as generating starting points and translating between programming languages. - AI-generated output is currently perceived as more reliable for developer workflows. - Designers may use AI to turn mockups into code, but much of design’s foundational work still involves: - Understanding user needs - Exploring problems nonlinearly - Learning about the broader problem space - AI initiatives increasingly originate outside design, with programmers, subject-matter experts, and stakeholders contributing ideas. ## AI Must Improve Collaboration, Not Just Individual Productivity - Eighty-five percent of respondents said AI had affected their personal productivity or workflows. - Common uses include text and image generation, brainstorming, and using AI as a sounding board or thought partner. - Respondents were three times more likely to report significant changes to individual workflows than to collaborative ones. - AI has not substantially changed group activities such as alignment or meeting facilitation. - Truly transformational AI products will need to support how teams work together, rather than focusing only on isolated tasks performed by individuals. ## Long-Term Effects Across Industries - Respondents in technology, professional and business services, and retail most often expected significant AI-driven changes to their products and services: - Technology: 41% - Professional and business services: 40% - Retail: 39% - Healthcare, energy and utilities, and telecommunications respondents expected the least impact over the following 12 months. - The report argues that realizing AI’s full potential requires considering how major institutions and essential services—not just software products—will evolve. ## Experimentation Before the Product Race - AI development is still characterized by experimentation, play, and research. - Teams face pressure to release new AI features quickly as new products, applications, and research appear constantly. - The recommended approach is to embrace uncertainty, iterate thoughtfully, and determine which ideas genuinely create value. - As the technology matures, the most successful products will likely come from careful exploration rather than simply adding AI features because of market hype.

Read original(opens in new tab)

Machine LearningAILarge Language ModelsGenerative AI+2

figma·Aug 3, 2023·2 min readCurated summary

Is your work style written in the stars? | Figma Blog

Astrology’s renewed popularity reflects people’s desire for recognition, guidance, and connection. Figma’s FigStrology widget brings those qualities into collaborative work by using machine learning to generate astrological personas and horoscopes from activity in FigJam files. The creators present it less as a serious predictive tool than as a playful way to encourage reflection and team interaction. ## Astrology as a Source of Connection - Astrology has existed for more than 4,000 years but was traditionally experienced privately. - Interest has increasingly shifted toward shared, collective experiences through apps such as CHANI and Co–Star. - Jacky Huang became interested during COVID after watching tarot readings on YouTube. - Huang says horoscopes appeal because they can make people feel recognized, understood, and seen. - Willy Wu views astrology as a source of belonging and reassurance in an uncertain world, even when its advice is humorous or not taken literally. ## Divination and Collaborative Design - Figma users are adapting divination practices for social and collaborative settings. - Danielle Baskin created Moonlight, a Figma-designed space for social tarot and reflection. - Figma’s persona quiz attracted more than 26,000 participants. - These projects position astrology and tarot as tools for conversation, self-reflection, and shared rituals rather than solely private practices. ## How FigStrology Was Created - Huang and Wu developed FigStrology as a FigJam widget. - The idea grew from Figma’s Photo Booth widget, which connects user behavior and activity with playful output. - FigStrology analyzes stickies and activity in a FigJam file, along with current “cosmic conditions.” - Machine learning then generates a unique persona and horoscope for the collaborative group or workspace. - The widget is intended as a lighthearted way to wind down a work session and make collaboration more engaging. ## Emphasizing Diversity and Team Strengths - The creators wanted FigStrology’s signs to highlight people and diversity. - Its premise is that individual strengths can complement other people’s weaknesses. - By turning team behavior into an imaginative horoscope, the widget encourages coworkers to notice different working styles and perspectives. Ultimately, FigStrology uses astrology as a playful collaboration device: not a replacement for genuine team communication, but a low-pressure way to spark reflection, humor, and connection at work.

Read original(opens in new tab)

Machine LearningFigmaFigjamCollaborative Design+1

figma·Jul 31, 2023·2 min readCurated summary

John Maeda on Creativity, AI, and the Human Pursuit of Uphill Thinking | Figma Blog

John Maeda argues that AI will transform creative work without eliminating the need for human creativity. While AI excels at efficiency, repetition, and finding the shortest path to an answer, meaningful creative breakthroughs often require experimentation, difficulty, and unconventional thinking. Designers should use AI to remove tedious work while deliberately preserving the human ability to pursue “uphill” paths. ## AI’s Promise and Threat to Creative Work - AI offers creatives greater efficiency and new capabilities, but may also produce large volumes of repetitive, cookie-cutter design. - Designers are learning to “speak machine” by collaborating with AI, training language models, and refining outputs to match their creative intentions. - Any technique developed with AI can potentially be replicated without its original creator, raising concerns about creative ownership and job security. ## Creativity Through Openness and Code - Maeda compares today’s AI concerns with his own experience in the 1990s, when he created distinctive algorithmic artwork. - Rather than keeping his methods secret, he open-sourced them at MIT so others could build their own creative work. - His involvement with MIT Scratch and Processing reflects his belief that programming can be a creative practice accessible to children, artists, and designers. ## Letting Computers Handle the Mundane - Computers are well suited to repetitive tasks such as: - Producing endless slide variations - Generating mockups - Performing image retouching - Automating this work could give designers more time to address emerging problems and develop more original ideas. - Maeda cites artist Jessie Shefrin’s observation that “by the time you come to the perfect solution, the problem has already changed,” emphasizing the need to keep moving rather than over-optimize a fixed answer. ## The Limits of Efficiency - AI is designed to find the shortest and most efficient route through a problem. - It can evaluate thousands or millions of possible paths rapidly and select the most efficient option. - Efficiency does not necessarily produce the most creative, meaningful, or impactful result. - The article presents “uphill thinking”—choosing difficult, indirect, or exploratory paths—as an important human strength in an increasingly automated world. Designers should embrace AI as a tool for reducing tedious labor, while continuing to protect the slower, less efficient processes that generate originality, insight, and meaningful creative work.

Read original(opens in new tab)

Machine LearningAIGenerative AICreative Coding+3

figma·Jun 20, 2023·3 min readCurated summary

AI: The Next Chapter in Design | Figma Blog

Figma argues that AI will become a core platform capability reshaping the entire product-development process, not merely another feature. It can accelerate ideation, design, and coding while allowing teams to focus more on problem-solving and creative judgment. Figma announced its acquisition of Diagram as part of this strategy, positioning AI as a force that will change how products are designed, what experiences are created, and who participates in the process. ## Figma’s AI strategy and Diagram acquisition - Figma acquired Diagram, founded by Jordan Singer, whose GPT-3-powered “Designer” plugin generated design concepts from simple prompts. - The acquisition brings Diagram’s team into Figma and builds on Figma’s existing investment in machine learning. - Figma’s open API has already enabled nearly 100 community-built AI plugins. - The company views AI as a platform underlying the entire product-development workflow. ## AI across the product-development process - During discovery, AI could: - Generate and synthesize early ideas from prompts. - Summarize discussions and concepts. - During design, AI could: - Use existing designs and design systems to provide recommendations. - Surface relevant components and patterns. - Help teams produce first drafts faster. - During development, AI could: - Infer design context more effectively. - Generate higher-quality, production-ready code. - The broader goal is to help teams do more work faster while moving their attention toward higher-level problem-solving. ## How design may evolve from pixels to patterns - Design systems already shifted designers away from repetitive details such as border radii and toward composition, direction, and judgment. - Atomic elements such as pixels became reusable components, enabling faster and more consistent workflows. - AI could extend this progression by generating higher-level structures and patterns. - Designers may focus less on assembling basic login components and more on inventing entirely new ways to authenticate. - AI might also recommend color palettes based on a project’s emotional tone or theme. - This could move design beyond familiar interfaces toward smoother, more intuitive, and more human experiences. ## What product teams may design - AI systems such as ChatGPT are shifting interaction away from navigating websites and apps toward asking questions and receiving answers. - AI can reduce the gap between a user’s intention and the actions required to achieve it. - For example, instead of opening a ride-hailing app, entering a destination, comparing options, and requesting a ride, a user could simply say, “Get me to JFK.” - Product builders will need to reconsider whether existing interfaces can deliver the same outcome with fewer steps and decisions. ## The changing role of designers - Technological change has historically transformed design without eliminating the need for thoughtful designers. - Designers have adapted to new platforms, collaborative workflows, and hybrid work. - Figma expects AI to change design roles and collaboration, but frames that shift as an opportunity to spend more time on creative direction, curation, and meaningful problem-solving. The practical recommendation is to treat AI as a foundational design and development capability rather than a standalone feature. Teams should explore how it can remove repetitive work while preserving human judgment, taste, and responsibility for the experiences they create.

Read original(opens in new tab)

Machine LearningAIDesign SystemsGenerative AI+4

figma·Mar 1, 2023·3 min readCurated summary

The Future of Design Systems is Automated | Figma Blog

Design systems are moving from static libraries toward automated, extensible ecosystems powered by plugins, widgets, and AI. These tools can automate repetitive work, expand Figma’s capabilities, and increasingly generate or recommend design solutions using existing system components. The article argues that automation will change designers’ responsibilities, but not eliminate the need for human judgment, creativity, and strategy. ## Plugins and Widgets as Design-System Extensions - Plugins have a long history in design and publishing software, dating back to tools such as HyperCard and QuarkXPress. - They created a broader ecosystem in which users could build and share custom effects, brushes, styles, and workflows. - In modern design systems, plugins generally serve two purposes: - Automating repetitive existing tasks. - Extending product capabilities through analytics, testing, accessibility checks, and other functionality. - Widgets add collaborative and visual tools directly to the design workspace, helping teams organize information and communicate around design systems. ## Automating Repetitive Tasks - Plugins can reduce manual work involved in maintaining and applying design-system assets. - Automation allows designers to spend less time on mechanical operations and more time on problem-solving and decision-making. - The broader trend reflects a shift from tools merely supporting designers to tools actively performing parts of the design process. ## Extending Design-System Capabilities - Plugins can provide capabilities that are not included in a core design application. - Examples include: - Gathering usage and library analytics. - Testing designs. - Improving accessibility. - Connecting design workflows to other tools and systems. - This extensibility enables teams to adapt their design environment to specialized organizational needs. ## AI-Assisted Design - Earlier experiments, such as Airbnb’s 2017 work on generating code from low-fidelity wireframes, demonstrated the potential of machine-learning-assisted design. - More recent tools such as Diagram’s Genius can analyze Figma files and suggest designs using components from an organization’s design system. - These developments suggest that AI is beginning to make earlier prototypes practical. - AI tools may eventually help generate interfaces, recommend components, complete workflows, or produce code from design input. ## Changing Roles and Responsibilities - Automation raises concerns about whether designers and developers will be replaced by software. - The article frames this as a question about how tools shape professional practice, rather than simply whether they eliminate jobs. - As routine production becomes automated, human designers may focus more on: - Defining problems. - Making judgments and trade-offs. - Establishing product direction. - Applying empathy, taste, and contextual understanding. - The future of design therefore depends on how practitioners adapt alongside increasingly capable tools. Design teams should treat plugins, widgets, and AI as ways to expand human capability rather than substitutes for design thinking. The most effective systems will combine automation for repetitive work with human oversight, creativity, and strategic judgment.

Read original(opens in new tab)

Machine LearningFigmaAIDesign Systems+4