A Berkeley View of Systems Challenges for AI

Abstract

With the increasing commoditization of computer vision, speech recognition and machine translation systems and the widespread deployment of learning-based back-end technologies such as digital advertising and intelligent infrastructures, AI (Artificial Intelligence) has moved from research labs to production. These changes have been made possible by unprecedented levels of data and computation, by methodological advances in machine learning, by innovations in systems software and architectures, and by the broad accessibility of these technologies. The next generation of AI systems promises to accelerate these developments and increasingly impact our lives via frequent interactions and making (often mission-critical) decisions on our behalf, often in highly personalized contexts. Realizing this promise, however, raises daunting challenges. In particular, we need AI systems that make timely and safe decisions in unpredictable environments, that are robust against sophisticated adversaries, and that can process ever increasing amounts of data across organizations and individuals without compromising confidentiality. These challenges will be exacerbated by the end of the Moore's Law, which will constrain the amount of data these technologies can store and process. In this paper, we propose several open research directions in systems, architectures, and security that can address these challenges and help unlock AI's potential to improve lives and society.

Berkeley Technical Report

References in corpus (7)

• Adam: A Method for Stochastic Optimization
• Deep Domain Confusion: Maximizing for Domain Invariance
• MXNet: A Flexible and Efficient Machine Learning Library for Heterogeneous Distributed Systems
• HOGWILD!: A Lock-Free Approach to Parallelizing Stochastic Gradient Descent
• Is feature selection secure against training data poisoning?
• Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World
• MLI: An API for Distributed Machine Learning

Cited by in corpus (40)

• Deep Reinforcement Learning: An Overview
• Slalom: Fast, Verifiable and Private Execution of Neural Networks in Trusted Hardware
• Deep learning of multi-element abundances from high-resolution spectroscopic data
• JALAD: Joint Accuracy- and Latency-Aware Deep Structure Decoupling for Edge-Cloud Execution
• Backdoor Attacks and Countermeasures on Deep Learning: A Comprehensive Review
• AI Safety Gridworlds
• Edge Intelligence: Paving the Last Mile of Artificial Intelligence with Edge Computing
• STRIP: A Defence Against Trojan Attacks on Deep Neural Networks
• Artificial Intelligence and its Role in Near Future
• Federated Learning via Over-the-Air Computation
• Machine Learning in Cyber-Security - Problems, Challenges and Data Sets
• Towards Ubiquitous AI in 6G with Federated Learning
• FutureMapping: The Computational Structure of Spatial AI Systems
• Stability-certified reinforcement learning: A control-theoretic perspective
• DeepHammer: Depleting the Intelligence of Deep Neural Networks through Targeted Chain of Bit Flips
• Safe Reinforcement Learning via Probabilistic Shields
• Skin Lesion Analysis Towards Melanoma Detection via End-to-end Deep Learning of Convolutional Neural Networks
• Orchestrating the Development Lifecycle of Machine Learning-Based IoT Applications: A Taxonomy and Survey
• Helen: Maliciously Secure Coopetitive Learning for Linear Models
• Large-Scale Discrete Fourier Transform on TPUs
• Shifting Capsule Networks from the Cloud to the Deep Edge
• MORPH-DSLAM: Model Order Reduction for PHysics-based Deformable SLAM
• PANDA: Facilitating Usable AI Development
• DeepPeep: Exploiting Design Ramifications to Decipher the Architecture of Compact DNNs
• Failure Prediction in Production Line Based on Federated Learning: An Empirical Study
• Distributing Intelligence to the Edge and Beyond
• A First Look at Deep Learning Apps on Smartphones
• Communication-Efficient Policy Gradient Methods for Distributed Reinforcement Learning
• Channel-Driven Monte Carlo Sampling for Bayesian Distributed Learning in Wireless Data Centers
• Fog Function: Serverless Fog Computing for Data Intensive IoT Services
• Device-Cloud Collaborative Learning for Recommendation
• Reaching Data Confidentiality and Model Accountability on the CalTrain
• Private Hierarchical Clustering and Efficient Approximation
• Scaling shared model governance via model splitting
• Towards Inference Delivery Networks: Distributing Machine Learning with Optimality Guarantees
• Machine Learning Systems for Intelligent Services in the IoT: A Survey
• Ten AI Stepping Stones for Cybersecurity
• Towards Edge-assisted Internet of Things: From Security and Efficiency Perspectives
• Verification of Uncertain POMDPs Using Barrier Certificates
• MyMigrationBot: A Cloud-based Facebook Social Chatbot for Migrant Populations