AI has a problem with objectifying women

🌈 Abstract

The article discusses the problematic objectification of women in the field of AI, particularly in the context of the recent ChatGPT 4.0 demo and the use of Scarlett Johansson's voice without her consent. It also examines broader issues of gender representation and bias in AI systems and demos.

🙋 Q&A

[01] The Objectification of Women in AI

1. What are the key issues discussed in the article regarding the objectification of women in AI?

• The article discusses how the small choices made in creating AI systems can have wide-ranging repercussions and contribute to entrenched perceptions and persistent stereotypes about women.
• It provides examples of how women's images and likenesses have been used without their consent in AI systems and demos, such as the use of Lenna Forsten's image as an industry standard and the practice of putting makeup on or changing the clothing of women in computer vision tasks.
• The article also discusses how the advent of increasingly realistic image generation models has led to the objectification of women, with AI-generated images of women often used to demonstrate the capabilities of these models.

2. What are the consequences of the objectification of women in AI, as discussed in the article?

• The article suggests that the objectification of women in AI has consequences not only for Hollywood celebrities like Scarlett Johansson, but also for "mere mortals" like the author.
• It argues that the situation with Scarlett Johansson is not the first and will not be the last instance of this kind of treatment of women and minorities by people in positions of power in AI.

3. What actions can be taken to address the objectification of women in AI, according to the article?

• For decision-makers, being more cognizant of issues of gender and power can translate into emphasizing diversity in positions of power, like boards and executive roles.
• For AI developers, this can be done by avoiding the implicit objectification of women via system demos, not using the common "traditionally attractive young nubile woman" images as the go-to example, and having explicit consent of those people chosen to illustrate the system.
• For the community at large, supporting organizations like Women in Machine Learning can help amplify the voices of women in the field and empower future generations.

[02] The Broader Context of Gender Representation and Bias in AI

1. What does the article say about the broader context of gender representation and bias in the field of AI?

• The article notes that the author has been working in AI for over a decade, in a field that has less than 12% of female researchers, and is often the only woman speaking on a panel or sitting at the table.
• The article suggests that making sure 50% of the world's population is represented in technology that has the potential to change humanity is truly exhausting.

2. How does the article connect the issues of objectification to the broader problem of gender representation and bias in AI?

• The article argues that the objectification of women in AI is part of a larger problem of gender representation and bias in the field, and that pushing back on this treatment and demanding respect, consent, and a seat at the table can help turn the tide on AI's longstanding tradition of objectifying women.