首页|Principles for the responsible application of Generative AI
Principles for the responsible application of Generative AI
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
Elsevier
The quest for Artificial Intelligence (AI) has comprised successive waves of excessive enthusiasm followed by long, dispirited lulls. Most recently, during the first 3-4 years of public access to Generative Artificial Intelligence (GenAI), many authors have bought into the bullish atmosphere, replaying consultancies' predictions about gold mines of process efficiency and innovation. A more balanced approach to the technology is needed. Instances of apparently positive results need calm analysis, firstly to distinguish mirages from genuine contributions; secondly, to identify ways to effectively exploit the new capabilities; and thirdly, to formulate guidance for the avoidance and mitigation of negative consequences. This article's first contribution is to ground the evaluation of GenAI's pathway, applications, impacts, implications and risks in a sufficiently deep appreciation of the technology's nature and key features. A wide range of sources is drawn on, in order to present descriptions of the processes involved in text-based GenAI. From those processes, 20 key characteristics are abstracted that together give rise to the promise and the threats GenAI embodies. The effects of GenAI derive not from the technological features alone, but also from the patterns within which it is put to use. By mapping usage patterns across to domains of application, the phenomenon's impacts and implications can be more reliably delineated. The analysis provides a platform whereby the article's final contribution can be made. Previously-formulated principles for the responsible application of AI of all kinds are applied in the particular context of GenAI.
Risk assessmentLarge language modelsGenAI artefactChatGPTRegulation
Roger Clarke
展开 >
Xamax Consultancy Pty Ltd, Canberra, Australia||Faculty of Law & Justice, University of New South Wales, Sydney, Australia||College of Systems & Society, Australian National University, Canberra, Australia
NETL
NSTL
Elsevier
