Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general.

/pdf/language-models-are-few-shot-learners-2fg8gvia7m.pdf

Language Models are Few-Shot Learners

Explainable Artificial Intelligence (XAI): Concepts, Taxonomies, Opportunities and Challenges toward Responsible AI

We propose a criterion for discrimination against a specified sensitive attribute in supervised learning, where the goal is to predict some target based on available features. Assuming data about the predictor, target, and membership in the protected group are available, we show how to optimally adjust any learned predictor so as to remove discrimination according to our definition. Our framework also improves incentives by shifting the cost of poor classification from disadvantaged groups to the decision maker, who can respond by improving the classification accuracy. We enourage readers to consult the more complete manuscript on the arXiv.

/pdf/equality-of-opportunity-in-supervised-learning-4ehzg3ysc9.pdf

Equality of opportunity in supervised learning

As machine learning systems become ubiquitous, there has been a surge of interest in interpretable machine learning: systems that provide explanation for their outputs. These explanations are often used to qualitatively assess other criteria such as safety or non-discrimination. However, despite the interest in interpretability, there is very little consensus on what interpretable machine learning is and how it should be measured. In this position paper, we first define interpretability and describe when interpretability is needed (and when it is not). Next, we suggest a taxonomy for rigorous evaluation and expose open questions towards a more rigorous science of interpretable machine learning.

Towards A Rigorous Science of Interpretable Machine Learning

Mobile applications are becoming increasingly ubiquitous and provide ever richer functionality on mobile devices. At the same time, such devices often enjoy strong connectivity with more powerful machines ranging from laptops and desktops to commercial clouds. This paper presents the design and implementation of CloneCloud, a system that automatically transforms mobile applications to benefit from the cloud. The system is a flexible application partitioner and execution runtime that enables unmodified mobile applications running in an application-level virtual machine to seamlessly off-load part of their execution from mobile devices onto device clones operating in a computational cloud. CloneCloud uses a combination of static analysis and dynamic profiling to partition applications automatically at a fine granularity while optimizing execution time and energy use for a target computation and communication environment. At runtime, the application partitioning is effected by migrating a thread from the mobile device at a chosen point to the clone in the cloud, executing there for the remainder of the partition, and re-integrating the migrated thread back to the mobile device. Our evaluation shows that CloneCloud can adapt application partitioning to different environments, and can help some applications achieve as much as a 20x execution speed-up and a 20-fold decrease of energy spent on the mobile device.

/pdf/clonecloud-elastic-execution-between-mobile-device-and-cloud-17imp9i7uj.pdf

CloneCloud: elastic execution between mobile device and cloud

Advocates of algorithmic techniques like data mining argue that these techniques eliminate human biases from the decision-making process. But an algorithm is only as good as the data it works with. Data is frequently imperfect in ways that allow these algorithms to inherit the prejudices of prior decision makers. In other cases, data may simply reflect the widespread biases that persist in society at large. In still others, data mining can discover surprisingly useful regularities that are really just preexisting patterns of exclusion and inequality. Unthinking reliance on data mining can deny historically disadvantaged and vulnerable groups full participation in society. Worse still, because the resulting discrimination is almost always an unintentional emergent property of the algorithm’s use rather than a conscious choice by its programmers, it can be unusually hard to identify the source of the problem or to explain it to a court.This Essay examines these concerns through the lens of American antidiscrimination law — more particularly, through Title VII’s prohibition of discrimination in employment. In the absence of a demonstrable intent to discriminate, the best doctrinal hope for data mining’s victims would seem to lie in disparate impact doctrine. Case law and the Equal Employment Opportunity Commission’s Uniform Guidelines, though, hold that a practice can be justified as a business necessity when its outcomes are predictive of future employment outcomes, and data mining is specifically designed to find such statistical correlations. Unless there is a reasonably practical way to demonstrate that these discoveries are spurious, Title VII would appear to bless its use, even though the correlations it discovers will often reflect historic patterns of prejudice, others’ discrimination against members of protected groups, or flaws in the underlying dataAddressing the sources of this unintentional discrimination and remedying the corresponding deficiencies in the law will be difficult technically, difficult legally, and difficult politically. There are a number of practical limits to what can be accomplished computationally. For example, when discrimination occurs because the data being mined is itself a result of past intentional discrimination, there is frequently no obvious method to adjust historical data to rid it of this taint. Corrective measures that alter the results of the data mining after it is complete would tread on legally and politically disputed terrain. These challenges for reform throw into stark relief the tension between the two major theories underlying antidiscrimination law: anticlassification and antisubordination. Finding a solution to big data’s disparate impact will require more than best efforts to stamp out prejudice and bias; it will require a wholesale reexamination of the meanings of “discrimination” and “fairness.”

Big Data's Disparate Impact

Big Data’s Disparate Impact

We survey 146 papers analyzing "bias" in NLP systems, finding that their motivations are often vague, inconsistent, and lacking in normative reasoning, despite the fact that analyzing "bias" is an inherently normative process. We further find that these papers' proposed quantitative techniques for measuring or mitigating "bias" are poorly matched to their motivations and do not engage with the relevant literature outside of NLP. Based on these findings, we describe the beginnings of a path forward by proposing three recommendations that should guide work analyzing "bias" in NLP systems. These recommendations rest on a greater recognition of the relationships between language and social hierarchies, encouraging researchers and practitioners to articulate their conceptualizations of "bias"---i.e., what kinds of system behaviors are harmful, in what ways, to whom, and why, as well as the normative reasoning underlying these statements---and to center work around the lived experiences of members of communities affected by NLP systems, while interrogating and reimagining the power relations between technologists and such communities.

Language (Technology) is Power: A Critical Survey of "Bias" in NLP

Online behavioral advertising (OBA) refers to the practice of tracking users across web sites in order to infer user interests and preferences. These interests and preferences are then used for selecting ads to present to the user. There is great concern that behavioral advertising in its present form infringes on user privacy. The resulting public debate — which includes consumer advocacy organizations, professional associations, and government agencies — is premised on the notion that OBA and privacy are inherently in conflict.In this paper we propose a practical architecture that enables targeting without compromising user privacy. Behavioral profiling and targeting in our system takes place in the user’s browser. We discuss the effectiveness of the system as well as potential social engineering and web-based attacks on the architecture. One complication is billing; ad-networks must bill the correct advertiser without knowing which ad was displayed to the user. We propose an efficient cryptographic billing system that directly solves the problem. We implemented the core targeting system as a Firefox extension and report on its effectiveness.

Adnostic: Privacy Preserving Targeted Advertising

We survey 146 papers analyzing "bias" in NLP systems, finding that their motivations are often vague, inconsistent, and lacking in normative reasoning, despite the fact that analyzing "bias" is an inherently normative process We further find that these papers' proposed quantitative techniques for measuring or mitigating "bias" are poorly matched to their motivations and do not engage with the relevant literature outside of NLP Based on these findings, we describe the beginnings of a path forward by proposing three recommendations that should guide work analyzing "bias" in NLP systems These recommendations rest on a greater recognition of the relationships between language and social hierarchies, encouraging researchers and practitioners to articulate their conceptualizations of "bias"---ie, what kinds of system behaviors are harmful, in what ways, to whom, and why, as well as the normative reasoning underlying these statements---and to center work around the lived experiences of members of communities affected by NLP systems, while interrogating and reimagining the power relations between technologists and such communities

/pdf/language-technology-is-power-a-critical-survey-of-bias-in-40xzysfmb0.pdf

Solon Barocas

Papers

Big Data's Disparate Impact

Big Data’s Disparate Impact

Language (Technology) is Power: A Critical Survey of "Bias" in NLP

Adnostic: Privacy Preserving Targeted Advertising

Language (Technology) is Power: A Critical Survey of "Bias" in NLP