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.

Australian experts have spoken out about a recent US study that claimed to show artificial intelligence can identify people with suicidal thoughts - by analysing their brain scans. It sounds promising - but it's worth pointing out only 79 people were studied, so are the results enough to show this is a path worth pursing? The research, published in Nature, studied brain activity in subjects when presented with a number of different words - like death, cruelty, trouble, carefree, good and praise. A machine-learning algorithm was then trained to see the nureal response differences between the two groups involved - those with suicidal thoughts, and those with non-suicidal thoughts. And it showed promise - the algorithm correctly identified 15 of 17 patients as belonging to the suicide group, and 16 of 17 healthy individuals as belonging to the control group.

Scientists were able to spot suicidal thoughts on a brain scan - paving the way to new methods for screening people at-risk of taking their own life. The groundbreaking study by Carnegie Mellon University showed unique brain activity in suicidal people when they heard death-related words. While suicide is the second-leading cause of death among young people in the US, it is notoriously difficult to predict. But experts say this unprecedented experiment offers never-before-seen insight into how suicidal people think, and what preventative measures could be taken. Using machine-learning algorithms, the team was able to identify how brain activity is affected by suicidal ideation and behavior, and how it compares to'healthy' brain activity.

Professor Scherer's team analysed the interviews using computer software that identified both verbal and non-verbal cues. Verbal content, such as mentioning death, repeated references to the past or heavy use of first-person pronouns, such as I, me and myself, were all common in the speech of suicidal patients. But what was surprising to researchers were the nonverbal cues. The found marked differences between the way suicidal and non-suicidal subjects spoke. Suicidal subjects had breathier speech, differences in pitch and other subtle changes in the tenseness or harshness of their voices, the experts wrote in the journal IEEE Transactions on Affective Computing.