Publications
Our teams aspire to make discoveries that impact everyone, and core to our approach is sharing our research and tools to fuel progress in the field.
Our teams aspire to make discoveries that impact everyone, and core to our approach is sharing our research and tools to fuel progress in the field.
Sort By
1 - 15 of 10033 publications
An intentional approach to managing bias in embedding models
Atilla P. Kiraly
Jungyeon Park
Rory Pilgrim
Charles Lau
Heather Cole-Lewis
Shravya Shetty
Krish Eswaran
Leo Anthony Celi
The Lancet Digital Health, 6 (2024), E126-E130
Preview abstract
Advances in machine learning for health care have brought concerns about bias from the research community; specifically, the introduction, perpetuation, or exacerbation of care disparities. Reinforcing these concerns is the finding that medical images often reveal signals about sensitive attributes in ways that are hard to pinpoint by both algorithms and people. This finding raises a question about how to best design general purpose pretrained embeddings (GPPEs, defined as embeddings meant to support a broad array of use cases) for building downstream models that are free from particular types of bias. The downstream model should be carefully evaluated for bias, and audited and improved as appropriate. However, in our view, well intentioned attempts to prevent the upstream components—GPPEs—from learning sensitive attributes can have unintended consequences on the downstream models. Despite producing a veneer of technical neutrality, the resultant end-to-end system might still be biased or poorly performing. We present reasons, by building on previously published data, to support the reasoning that GPPEs should ideally contain as much information as the original data contain, and highlight the perils of trying to remove sensitive attributes from a GPPE. We also emphasise that downstream prediction models trained for specific tasks and settings, whether developed using GPPEs or not, should be carefully designed and evaluated to avoid bias that makes models vulnerable to issues such as distributional shift. These evaluations should be done by a diverse team, including social scientists, on a diverse cohort representing the full breadth of the patient population for which the final model is intended.
View details
Limoncello: Prefetchers for Scale
Carlos Villavieja
Baris Kasikci
Proceedings of the 28th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Association for Computing Machinery, New York, NY, United States (2024)
Preview abstract
This paper presents Limoncello, a novel software system that dynamically configures data prefetching for high utilization systems. We demonstrate that in resource-constrained environments, such as large data centers, traditional methods of hardware prefetching can increase memory latency and decrease available memory bandwidth. To address this, Limoncello dynamically configures data prefetching, disabling hardware prefetchers when memory bandwidth utilization is high and leveraging targeted software prefetching to reduce cache misses when hardware prefetchers are disabled. Limoncello is software-centric and does not require any modifications to hardware. Our evaluation of the deployment on a real-world hyperscale system reveals that Limoncello unlocks significant performance gains for high-utilization systems: it improves application throughput by 10%, due to a 15% reduction in memory latency, while maintaining minimal change in cache miss rate for targeted library functions.
View details
Preview abstract
Proving mathematical theorems at the olympiad level represents a notable milestone in human-level automated reasoning, owing to their reputed difficulty among the world’s best talents in pre-university mathematics. Current machine-learning approaches, however, are not applicable to most mathematical domains owing to the high cost of translating human proofs into machine-verifiable format. The problem is even worse for geometry because of its unique translation challenges, resulting in severe scarcity of training data. We propose AlphaGeometry, a theorem prover for Euclidean plane geometry that sidesteps the need for human demonstrations by synthesizing millions of theorems and proofs across different levels of complexity. AlphaGeometry is a neuro-symbolic system that uses a neural language model, trained from scratch on our large-scale synthetic data, to guide a symbolic deduction engine through infinite branching points in challenging problems. On a test set of 30 latest olympiad-level problems, AlphaGeometry solves 25, outperforming the previous best method that only solves ten problems and approaching the performance of an average International Mathematical Olympiad (IMO) gold medallist. Notably, AlphaGeometry produces human-readable proofs, solves all geometry problems in the IMO 2000 and 2015 under human expert evaluation and discovers a generalized version of a translated IMO theorem in 2004.
View details
UINav: A Practical Approach to Train On-Device Automation Agents
Wei Li
Fu-Lin Hsu
Will Bishop
Folawiyo Campbell-Ajala
2024 Annual Conference of the North American Chapter of the Association for Computational Linguistics (NAACL 2024) - Industry Track
Preview abstract
Automation systems that can autonomously drive application user interfaces to complete user tasks are of great benefit, especially when users are situationally or permanently impaired. Prior automation systems do not produce generalizable models while AI-based automation agents work reliably only in simple, hand-crafted applications or incur high computation costs. We propose UINav, a demonstration-based approach to train automation agents that fit mobile devices, yet achieving high success rates with modest numbers of demonstrations. To reduce the demonstration overhead, UINav, uses a referee model that provides users with immediate feedback on tasks where the agent fails, and automatically augments human demonstrations to increase diversity in training data. Our evaluation shows that with only 10 demonstrations UINav, can achieve 70% accuracy, and that with enough demonstrations it can surpass 90% accuracy.
View details
Efficient Location Sampling Algorithms for Road Networks
Vivek Kumar
Santhoshini Velusamy
WebConf (2024)
Preview abstract
Many geographic information systems applications rely on the data provided by user devices in the road network. Such applications include traffic monitoring, driving navigation, detecting road closures or the construction of new roads, etc. This signal is collected by sampling locations from the user trajectories and is a critical process for all such systems. Yet, it has not been sufficiently studied in the literature. The most natural way to sample a trajectory is perhaps using a frequency based algorithm, e.g., sample every $x$ seconds. However, as we argue in this paper, such a simple strategy can be very wasteful in terms of resources (e.g., server-side processing, user battery) and in terms of the amount of user data that it maintains. In this work we conduct a horizontal study of various location sampling algorithms (including frequency-based, road geography-based, reservoir-sampling based, etc.) and extract their trade-offs in terms of various metrics of interest, such as, the size of the stored data and the induced quality of training for prediction tasks (e.g., predicting speeds) using the road network of New York City.
View details
Large Language Models are Effective Text Rankers with Pairwise Ranking Prompting
Conference of the North American Chapter of the Association for Computational Linguistics (NAACL) (2024)
Preview abstract
Ranking documents using Large Language Models (LLMs) by directly feeding the query and candidate documents into the prompt is an interesting and practical problem. However, researchers have found it difficult to outperform fine-tuned baseline rankers on benchmark datasets. We analyze pointwise and listwise ranking prompts used by existing methods and argue that off-the-shelf LLMs do not fully understand these challenging ranking formulations. In this paper, we propose to significantly reduce the burden on LLMs by using a new technique called Pairwise Ranking Prompting (PRP). Our results are the first in the literature to achieve state-of-the-art ranking performance on standard benchmarks using moderate-sized open-sourced LLMs. On TREC-DL 2019&2020, PRP based on the Flan-UL2 model with 20B parameters performs favorably with the previous best approach in the literature, which is based on the blackbox commercial GPT-4 that has 50x (estimated) model size, while outperforming other LLM-based solutions, such as InstructGPT which has 175B parameters, by over 10% for all ranking metrics. By using the same prompt template on seven BEIR tasks, PRP outperforms supervised baselines and outperforms the blackbox commercial ChatGPT solution by 4.2% and pointwise LLM-based solutions by more than 10% on average NDCG@10. Furthermore, we propose several variants of PRP to improve efficiency and show that it is possible to achieve competitive results even with linear complexity.
View details
A Toolbox for Surfacing Health Equity Harms and Biases in Large Language Models
Heather Cole-Lewis
Nenad Tomašev
Liam McCoy
Leo Anthony Celi
Alanna Walton
Akeiylah DeWitt
Philip Mansfield
Sushant Prakash
Joelle Barral
Ivor Horn
Karan Singhal
Arxiv (2024) (to appear)
Preview abstract
Large language models (LLMs) hold immense promise to serve complex health information needs but also have the potential to introduce harm and exacerbate health disparities. Reliably evaluating equity-related model failures is a critical step toward developing systems that promote health equity. In this work, we present resources and methodologies for surfacing biases with potential to precipitate equity-related harms in long-form, LLM-generated answers to medical questions and then conduct an empirical case study with Med-PaLM 2, resulting in the largest human evaluation study in this area to date. Our contributions include a multifactorial framework for human assessment of LLM-generated answers for biases, and EquityMedQA, a collection of seven newly-released datasets comprising both manually-curated and LLM-generated questions enriched for adversarial queries. Both our human assessment framework and dataset design process are grounded in an iterative participatory approach and review of possible biases in Med-PaLM 2 answers to adversarial queries. Through our empirical study, we find that the use of a collection of datasets curated through a variety of methodologies, coupled with a thorough evaluation protocol that leverages multiple assessment rubric designs and diverse rater groups, surfaces biases that may be missed via narrower evaluation approaches. Our experience underscores the importance of using diverse assessment methodologies and involving raters of varying backgrounds and expertise. We emphasize that while our framework can identify specific forms of bias, it is not sufficient to holistically assess whether the deployment of an AI system promotes equitable health outcomes. We hope the broader community leverages and builds on these tools and methods towards realizing a shared goal of LLMs that promote accessible and equitable healthcare for all.
View details
Resolving Code Review Comments with Machine Learning
Alexander Frömmgen
Peter Choy
Elena Khrapko
Marcus Revaj
2024 IEEE/ACM 46th International Conference on Software Engineering: Software Engineering in Practice (ICSE-SEIP) (to appear)
Preview abstract
Code reviews are a critical part of the software development process, taking a significant amount of the code authors’ and the code reviewers’ time. As part of this process, the reviewer inspects the proposed code and asks the author for code changes through comments written in natural language. At Google, we see millions of reviewer comments per year, and authors require an average of ∼60 minutes active shepherding time between sending changes for review and finally submitting the change. In our measurements, the required active work time that the code author must devote to address reviewer comments grows almost linearly with the number of comments. However, with machine learning (ML), we have an opportunity to automate and streamline the code-review process, e.g., by proposing code changes based on a comment’s text.
We describe our application of recent advances in large sequence models in a real-world setting to automatically resolve code-review comments in the day-to-day development workflow at Google. We present the evolution of this feature from an asynchronous generation of suggested edits after the reviewer sends feedback, to an interactive experience that suggests code edits to the reviewer at review time. In deployment, code-change authors at Google address 7.5% of all reviewer comments by applying an ML-suggested edit. The impact of this will be to reduce the time spent on code reviews by hundreds of thousands of engineer hours annually at Google scale. Unsolicited, very positive feedback highlights that the impact of ML-suggested code edits increases Googlers’ productivity and allows them to focus on more creative and complex tasks.
View details
Preview abstract
We present XDTK, an open-source Unity/Android toolkit for prototyping multi-device interactions in extended reality (XR). With the Unity package and Android app provided in XDTK, data from any number of devices (phones, tablets, or wearables) can be streamed to and surfaced within a Unity-based XR application. ARCore-supported device also provide self-tracked pose data. Devices on the same local network are automatically discovered by the Unity server and their inputs are routed using a custom event framework. We designed XDTK to be modular and easily extendable to enable fast, simple, and effective prototyping of multi-device experiences by both researchers and developers.
View details
A Decentralized SDN Architecture for the WAN
Nitika Saran
Ashok Narayanan
Sylvia Ratnasamy
Ankit Singla
Hakim Weatherspoon
2024 ACM Special Interest Group on Data Communication (SIGCOMM) (2024)
Preview abstract
Motivated by our experiences operating a global WAN, we argue that SDN’s reliance on infrastructure external to the data plane has significantly complicated the challenge of maintaining high availability. We propose a new decentralized SDN (dSDN) architecture in which SDN control logic instead runs within routers, eliminating the control plane’s reliance on external infrastructure and restoring fate sharing between control and data planes.
We present dSDN as a simpler approach to realizing the benefits of SDN in the WAN. Despite its much simpler design, we show that dSDN is practical from an implementation viewpoint, and outperforms centralized SDN in terms of routing convergence and SLO impact.
View details
AI-Enhanced API Design: A New Paradigm in Usability and Efficiency
Mak Ahmad
David R Karger
Kwan-Liu Ma
CHI EA '24: Extended Abstracts of the 2024 CHI Conference on Human Factors in Computing Systems (2024)
Preview abstract
This study uses mixed methods to evaluate API design methods, focusing on design and consumption phases. Our goal was to understand the impact of API governance approaches on productivity and usability. A controlled developer experiment (n=34) demonstrated
a 10% increased requirement fulfillment using API Improvement Proposals (AIPs) and linter versus no protocols. Meanwhile, 73% of 33 surveyed API consumers preferred AIP-aligned designs for enhanced usability and comprehensibility. Complementing this, a
custom large language model called the API Architect received average expert ratings of just 5/10 for specification quality, revealing gaps versus manual design. The quantitative performance metrics combined with qualitative user feedback provide evidence from
multiple angles that strategically integrating industry best practices with maturing AI capabilities can meaningfully improve API design outcomes. This research offers empirical insights from developer and consumer perspectives to advance scholarly discourse
and industry practice regarding optimal API design workflows.
View details
Preview abstract
Graphs are a powerful tool for representing and analyzing complex relationships in real-world applications such as social networks, recommender systems, and computational finance. Reasoning on graphs is essential for drawing inferences about the relationships between entities in a complex system, and to identify hidden patterns and trends. Despite the remarkable progress in automated reasoning with natural text, reasoning on graphs with large language models (LLMs) remains an understudied problem. In this work, we perform the first comprehensive study of encoding graph-structured data as text for consumption by LLMs. We show that LLM performance on graph reasoning tasks varies on three fundamental levels: (1) the graph encoding method, (2) the nature of the graph task itself, and (3) interestingly, the very structure of the graph considered. These novel results provide valuable insight on strategies for encoding graphs as text. Using these insights we illustrate how the correct choice of encoders can boost performance on graph reasoning tasks inside LLMs by 4.8% to 61.8%, depending on the task.
View details
Conversational AI in health: Design considerations from a Wizard-of-Oz dermatology case study with users, clinicians and a medical LLM
Brenna Li
Amy Wang
Patricia Strachan
Julie Anne Seguin
Sami Lachgar
Karyn Schroeder
Renee Wong
Extended Abstracts of the 2024 CHI Conference on Human Factors in Computing Systems, Association for Computing Machinery, pp. 10
Preview abstract
Although skin concerns are common, access to specialist care is limited. Artificial intelligence (AI)-assisted tools to support medical decisions may provide patients with feedback on their concerns while also helping ensure the most urgent cases are routed to dermatologists. Although AI-based conversational agents have been explored recently, how they are perceived by patients and clinicians is not well understood. We conducted a Wizard-of-Oz study involving 18 participants with real skin concerns. Participants were randomly assigned to interact with either a clinician agent (portrayed by a dermatologist) or an LLM agent (supervised by a dermatologist) via synchronous multimodal chat. In both conditions, participants found the conversation to be helpful in understanding their medical situation and alleviate their concerns. Through qualitative coding of the conversation transcripts, we provide insight on the importance of empathy and effective information-seeking. We conclude with design considerations for future AI-based conversational agents in healthcare settings.
View details
Preview abstract
The evolution of AI is a pivotal moment in history, but it’s not the first time we have experienced technological advances that have changed how humans work. By looking at the advances in automobiles, we are reminded of the importance of focusing on our developers' needs and goals.
View details
Preview abstract
2022 marked the 50th anniversary of memory safety vulnerabilities, first reported by Anderson et al. Half a century later, we are still dealing with memory safety bugs despite substantial investments to improve memory unsafe languages.
Like others', Google’s data and internal vulnerability research show that memory safety bugs are widespread and one of the leading causes of vulnerabilities in memory-unsafe codebases. Those vulnerabilities endanger end users, our industry, and the broader society.
At Google, we have decades of experience addressing, at scale, large classes of vulnerabilities that were once similarly prevalent as memory safety issues. Based on this experience we expect that high assurance memory safety can only be achieved via a Secure-by-Design approach centered around comprehensive adoption of languages with rigorous memory safety guarantees.
We see no realistic path for an evolution of C++ into a language with rigorous memory safety guarantees that include temporal safety. As a consequence, we are considering a gradual transition of C++ code at Google towards other languages that are memory safe.
Given the large volume of pre-existing C++, we believe it is nonetheless necessary to improve the safety of C++ to the extent practicable. We are considering transitioning to a safer C++ subset, augmented with hardware security features like MTE.
View details