Left to right: Rachel Coldicutt, David Leslie, Rumman Chowdhury, Noura Al Moubayed, Wendy Hall.
Royal Society/Debbie Rowe
On the second day of the Women and the Future of Science conference at the Royal Society in London, I encountered significant challenges with AI transcription software. It consistently mistyped names, which strained my ability to focus on the impactful discussions surrounding artificial intelligence, particularly concerning the erasure of women in contemporary AI technologies.
This issue extends beyond the well-documented bias in AI algorithms stemming from training datasets that often lack gender diversity.
Sessions led by renowned computer scientists, including Wendy Hall, aim to tackle a pressing concern: the predominance of male designers in crafting transformative AI technologies that greatly impact society.
Historically, technology has been a male-dominated domain, with current statistics showing that only 25 percent of computer science students in the UK are women. In recent years, Silicon Valley’s environment has become increasingly hostile towards women, particularly as generative AI continues to evolve.
“There has been a significant setback over the last two years,” states David Leslie, Director of Ethics and Responsible Innovation Research at the Alan Turing Institute. “Debates regarding the generational damage inflicted on women in science by the Trump administration are not merely speculative; we’re regressing.”
Last year, President Donald Trump issued an executive order that targeted the concept of “woke AI,” urging the US National Institute of Standards and Technology to re-evaluate its AI risk management framework, stripping away references to misinformation, diversity, equity, inclusion, and climate change.
Among the panelists was Rumman Chowdhury, a data scientist and former special envoy for AI science, who previously oversaw ethics and accountability at Twitter under Elon Musk before being dismissed along with her team. She highlights that the notion of woke AI emerged from sexist attitudes within Silicon Valley prior to the President’s directives.
When asked to envision AI devoid of female contributions, panelists noted that we are already witnessing this reality. “In the sphere of frontier AI, we are indeed in an AI landscape without women,” declares Chaudhry, while Rachel Coldicutt emphasizes that lacking women in AI is not a distant fantasy, but a current reality.
The implications are profound. From crash test dummies to medical research, a longstanding trend exists where technology is built with male bodies and needs in mind, a phenomenon termed the gender data gap. The ramifications of this gap can range from inconvenient to life-threatening.
AI’s influence will permeate various aspects of life, including employment, education, and healthcare. However, as highlighted by Chaudhry, women currently receive only 2% of venture capital funding, and less than 1 percent of healthcare research funding addresses women’s health. “We must utilize technology for everyone, not just the elite,” Coldicutt stressed.
What actions should be taken? Coldicutt argues that existing AI models are crippled by centuries of bias, making rectification nearly impossible. “We need alternative models,” she insists, emphasizing the importance of fostering systems that prioritize care for both people and the planet.
Chaudhry, a co-founder of the nonprofit Humane Intelligence, which aids companies in enhancing accountability and fairness in AI systems, notes that much of current AI development is driven by a misplaced urgency focused on existential threats to jobs and humanity. “If your house were on fire, you wouldn’t contemplate your mother’s jewelry in that moment,” she explains. This sense of urgency leads to the neglect of essential factors, including diversity.
For the upcoming generation, Leslie advocates that to empower youth in developing AI for social benefit, we must reevaluate the economic and political frameworks surrounding AI development. “We need to begin by redefining incentives.”
Ultimately, we may need to redefine the very notion of intelligence in the context of AI to embrace a wider, more diverse array of perspectives. Much of the foundational thought on AI, including its definitions, arose from a landmark conference held at Dartmouth College in the 1950s—an event composed entirely of men, as Hall points out.
Topics:
Source: www.newscientist.com
