Structured access: an emerging paradigm for safe AI deployment

Shevlane, Toby

doi:10.48550/arxiv.2201.05159

Cited by 5 publications

(4 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Another objective would be preventing users from circumventing a model's restrictions to modify or reproduce it. In that regard, Shevlane [39] proposes structured access as an emerging paradigm that constructs a controlled, arm's length interaction between an AI system and its user.…”

Section: Theoretical Proposals To Regulate Frontier Ai Modelsmentioning

confidence: 99%

The EU AI Act: A pioneering effort to regulate frontier AI?

Bas,

Salinas,

Tinoco

et al. 2024

View full text Add to dashboard Cite

The emergence of increasingly capable artificial intelligence (AI) systems has raised concerns about the potential extreme risks associated with them. The issue has drawn substantial attention in academic literature and compelled legislators of regulatory frameworks like the European Union AI Act (AIA) to readapt them to the new paradigm. This paper examines whether the European Parliament’s draft of the AIA constitutes an appropriate approach to address the risks derived from frontier models. In particular, we discuss whether the AIA reflects the policy needs diagnosed by recent literature and determine if the requirements falling on providers of foundation models are appropriate, sufficient, and durable. We find that the provisions are generally adequate, but insufficiently defined in some areas and lacking in others. Finally, the AIA is characterized as an evolving framework whose durability will depend on the institutions’ ability to adapt to future progress.

show abstract

Section: Theoretical Proposals To Regulate Frontier Ai Modelsmentioning

confidence: 99%

The EU AI Act: A pioneering effort to regulate frontier AI?

Bas,

Salinas,

Tinoco

et al. 2024

View full text Add to dashboard Cite

show abstract

“…Consistent with the recommendations of , we believe research access to publicly benchmark and document these models is necessary, even if the broader practices for model release will differ across model providers. To this end, we recommend patterns of developer-mediated access as potential middlegrounds to ensure these models can be benchmarked transparently as a form of structured model access (Shevlane, 2022).…”

Section: Missing Modelsmentioning

confidence: 99%

Holistic Evaluation of Language Models

Liang¹,

Bommasani²,

Lee³

et al. 2022

Preprint

View full text Add to dashboard Cite

Language models (LMs) are becoming the foundation for almost all major language technologies, but their capabilities, limitations, and risks are not well understood. We present Holistic Evaluation of Language Models (HELM) to improve the transparency of language models. First, we taxonomize the vast space of potential scenarios (i.e. use cases) and metrics (i.e. desiderata) that are of interest for LMs. Then we select a broad subset based on coverage and feasibility, noting what's missing or underrepresented (e.g. question answering for neglected English dialects, metrics for trustworthiness). Second, we adopt a multi-metric approach: We measure 7 metrics (accuracy, calibration, robustness, fairness, bias, toxicity, and efficiency) for each of 16 core scenarios to the extent possible (87.5% of the time), ensuring that metrics beyond accuracy don't fall to the wayside, and that trade-offs across models and metrics are clearly exposed. We also perform 7 targeted evaluations, based on 26 targeted scenarios, to more deeply analyze specific aspects (e.g. knowledge, reasoning, memorization/copyright, disinformation). Third, we conduct a large-scale evaluation of 30 prominent language models (spanning open, limited-access, and closed models) on all 42 scenarios, including 21 scenarios that were not previously used in mainstream LM evaluation. Prior to HELM, models on average were evaluated on just 17.9% of the core HELM scenarios, with some prominent models not sharing a single scenario in common. We improve this to 96.0%: now all 30 models have been densely benchmarked on a set of core scenarios and metrics under standardized conditions. Our evaluation surfaces 25 top-level findings concerning the interplay between different scenarios, metrics, and models. For full transparency, we release all raw model prompts and completions publicly 3 for further analysis, as well as a general modular toolkit for easily adding new scenarios, models, metrics, and prompting strategies. 4 We intend for HELM to be a living benchmark for the community, continuously updated with new scenarios, metrics, and models.

show abstract

“…Additionally, for many contemporary LLMs, there are other individuation choices which must be made. For example, model developers (especially commercial developers) often provide 'structured access' to models (Shevlane, 2022), either via an API or a web application. This means that the function p θ underlying these models is always evaluated with not only a tokenizer and inference procedure, but various additional elements.…”

Section: Models and Background Conditions: Using Llms As A Case Studymentioning

confidence: 99%

Is Deontic Evaluation Capable of Doing What it is For?

Sharadin

Greve²

2021

JESP

View full text Add to dashboard Cite

Many philosophers think the distinctive function of deontic evaluation is to guide action. This idea is used in arguments for a range of substantive claims. In this paper, we entirely do one completely destructive thing and partly do one not entirely constructive thing. The first thing: we argue that there is an unrecognized gap between the claim that the function of deontic evaluation is to guide action and attempts to put that claim to use. We consider and reject four arguments intended to bridge this gap. The interim conclusion is thus that arguments starting with the claim that the function of deontic evaluation is to guide action have a lacuna. The second thing: we consider a different tack for making arguments of this sort work. We sketch a methodology one could accept that would do the trick. Unfortunately, as we’ll explain, although this methodology would bridge the gap in arguments that put claims about the function of deontic evaluation to work, it would do so in a way that vitiates any interest we might have in such arguments. As an aside, we’ll also point out how epistemologists, who have recently become interested in the function of epistemic evaluation, appear to already recognize this fact. The conclusion is hence a dilemma: either arguments from deontic function to substance have a lacuna or such arguments lack teeth.

show abstract

Structured access: an emerging paradigm for safe AI deployment

Cited by 5 publications

References 0 publications

The EU AI Act: A pioneering effort to regulate frontier AI?

The EU AI Act: A pioneering effort to regulate frontier AI?

Holistic Evaluation of Language Models

Is Deontic Evaluation Capable of Doing What it is For?

Contact Info

Product

Resources

About