An official website of the United States government
Abstract This paper presents a novel application of convolutional neural network (CNN) models for filtering the intraseasonal variability of the tropical atmosphere. In this deep learning filter, two convolutional layers are applied sequentially in a supervised machine learning framework to extract the intraseasonal signal from the total daily anomalies. The CNN-based filter can be tailored for each field similarly to fast Fourier transform filtering methods. When applied to two different fields (zonal wind stress and outgoing longwave radiation), the index of agreement between the filtered signal obtained using the CNN-based filter and a conventional weight-based filter is between 95% and 99%. The advantage of the CNN-based filter over the conventional filters is its applicability to time series with the length comparable to the period of the signal being extracted. Significance StatementThis study proposes a new method for discovering hidden connections in data representative of tropical atmosphere variability. The method makes use of an artificial intelligence (AI) algorithm that combines a mathematical operation known as convolution with a mathematical model built to reflect the behavior of the human brain known as artificial neural network. Our results show that the filtered data produced by the AI-based method are consistent with the results obtained using conventional mathematical algorithms. The advantage of the AI-based method is that it can be applied to cases for which the conventional methods have limitations, such as forecast (hindcast) data or real-time monitoring of tropical variability in the 20–100-day range.
more »
« less
Seeing Seeds Beyond Weeds: Green Teaming Generative AI for Beneficial Uses
Large generative AI models (GMs) like GPT and DALL-E are trained to generate content for general, wide-ranging purposes. GM content filters are generalized to filter out content which has a risk of harm in many cases, e.g., hate speech. However, prohibited content is not always harmful -- there are instances where generating prohibited content can be beneficial. So, when GMs filter out content, they preclude beneficial use cases along with harmful ones. Which use cases are precluded reflects the values embedded in GM content filtering. Recent work on red teaming proposes methods to bypass GM content filters to generate harmful content. We coin the term green teaming to describe methods of bypassing GM content filters to design for beneficial use cases. We showcase green teaming by: 1) Using ChatGPT as a virtual patient to simulate a person experiencing suicidal ideation, for suicide support training; 2) Using Codex to intentionally generate buggy solutions to train students on debugging; and 3) Examining an Instagram page using Midjourney to generate images of anti-LGBTQ+ politicians in drag. Finally, we discuss how our use cases demonstrate green teaming as both a practical design method and a mode of critique, which problematizes and subverts current understandings of harms and values in generative AI.
more »
« less
- PAR ID:
- 10469270
- Publisher / Repository:
- ICML Workshop
- Date Published:
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
In response to growing environmental concerns regarding the presence of per- and polyfluoroalkyl substances (PFAS) in landfills, this study explores PFAS permeation through pinhole defects of high-density polyethylene (HDPE) geomembranes (GMs) experimentally. Specifically, this study aims to: (i) investigate the adsorption of PFAS onto HDPE GMs, (ii) evaluate the effectiveness of GMs experimentally in retaining PFAS-laden leachate in the event of a puncture failure, (iii) assess the critical conditions leading to puncture failure of GM using mechanical characterization testing with complementary finite element method (FEM) analyses with the input data from mechanical characterization. Our findings show limited intermolecular attractive interactions between PFAS and GMs, and surfactant properties of PFAS contribute to higher leachate permeation through pinholes. In general, highly fluorinated, short chain PFAS exhibit increased permeation rates, which was attributed to their size and greater propensity to align at the water-air interface. This study underlines the environmental implications of PFAS-laden leachates especially when there are no proper liner systems or leachate collection systems in place underscoring the necessity for modern landfill design and management practices to mitigate environmental risks associated with PFAS.more » « less
-
Abstract A critical task to better quantify changes in precipitation (P) mean and extreme statistics due to global warming is to gain insights into the underlying physical generating mechanisms (GMs). Here, the dominant GMs associated with daily P recorded at 2861 gauges in the Conterminous United States from 1980 to 2018 were identified from atmospheric reanalyses and publicly available datasets. The GMs include fronts (FRT), extratropical cyclones (ETC), atmospheric rivers (AR), tropical cyclones (TC), and North American Monsoon (NAM). Climatologies of the GM occurrences were developed for the nonzero P (NZP) and annual P maxima (APM) samples, characterizing the marginal and extreme P distributions, respectively. FRT is everywhere the most frequent (45-75%) GM of NZP followed by ETC (12-33%). The FRT contribution declines for APM (19-66%), which are dominated by AR (50-65%) in western regions and affected by TC (10-18%) in southern and eastern regions. The GM frequencies exhibit trends with the same signs over large regions, which are not statistically significant except for an increase in FRT (TC) frequency in the Northeast (central region). Two-sample tests showed well-defined spatial patterns with regions where (1) both the marginal and extreme P distributions of the two dominant GMs likely belong to different statistical populations, and (2) only the marginal or the extreme distributions could be considered statistically different. These results were interpreted throughL-moments and parametric distributions that adequately model NZP and APM frequency. This work provides useful insights to incorporate mixed populations and nonstationarity in P frequency analyses.more » « less
-
Harmful textual content is pervasive on social media, poisoning online communities and negatively impacting participation. A common approach to this issue is developing detection models that rely on human annotations. However, the tasks required to build such models expose annotators to harmful and offensive content and may require significant time and cost to complete. Generative AI models have the potential to understand and detect harmful textual content. We used ChatGPT to investigate this potential and compared its performance with MTurker annotations for three frequently discussed concepts related to harmful textual content on social media: Hateful, Offensive, and Toxic (HOT). We designed five prompts to interact with ChatGPT and conducted four experiments eliciting HOT classifications. Our results show that ChatGPT can achieve an accuracy of approximately 80% when compared to MTurker annotations. Specifically, the model displays a more consistent classification for non-HOT comments than HOT comments compared to human annotations. Our findings also suggest that ChatGPT classifications align with the provided HOT definitions. However, ChatGPT classifies “hateful” and “offensive” as subsets of “toxic.” Moreover, the choice of prompts used to interact with ChatGPT impacts its performance. Based on these insights, our study provides several meaningful implications for employing ChatGPT to detect HOT content, particularly regarding the reliability and consistency of its performance, its understanding and reasoning of the HOT concept, and the impact of prompts on its performance. Overall, our study provides guidance on the potential of using generative AI models for moderating large volumes of user-generated textual content on social media.more » « less
-
Design thinking emphasizes that in addition to being creative, design solutions should be empathetic. Yet, research suggests there may be a tension between these goals, where focusing on empathy comes at a cost to creativity, sometimes by inducing fixation. We investigated this phenomenon through a quasi-experimental design with novice designers, contrasting two structured ideation techniques in which participants (N = 47) generated bad ideas prior to proposing beneficial ideas. Specifically, they used the wrong theory protocol (WTP) to generate harmful and humiliating ideas, and a variant in which they instead generated silly and impossible ideas (SIP). We used qualitative analysis to characterize their bad and beneficial ideas. Across two realistic design challenges, we found students’ initial bad design work was shaped by the technique they used, and that those who generated humiliating ideas were more likely to generate empathetic beneficial ideas afterward. No systematic differences were found in the breadth of solution ideas, suggesting this technique does not come at a cost to creativity. As a quick and easy-to-use technique, generating humiliating ideas prior to generating beneficial ideas holds promise as a means to reach design solutions that are both empathetic and creative.more » « less
