"You believe what this flirty OpenAI chatbot says to women in tech!"

 Besides Sky ChatGPT's similarity to Scarlett Johansson's voice, companies building AI for humanity have a problem when very few of their researchers are women, says Parmy Olson.

Source : Google.com/images.app.goo.gl/hgKutswwe7pVajc1A

It's hard not to cringe at the demo of ChatGPT's latest update. Instead of presenting a more reliable-sounding chatbot, OpenAI gave the world a chatbot perfectly suited to submissive women, giggling at the antics of its male researchers, and complimenting their outfits.

The outrage raised over the voice's similarity to Scarlett Johansson misses the deeper point. The world's leading AI developers are creating software that reinforces stereotypes about women. And there's a big reason: There are too few of those involved.

At OpenAI, only 18 percent of staff working on its technology development are women, according to data compiled for Bloomberg Opinion by business intelligence firm Glass.ai, which uses machine learning technology to scrutinize tech company websites and thousands of AI-focused employee LinkedIn profiles. . The creator of ChatGPT was the worst performer among other leading companies in a survey conducted in May.

Although OpenAI's chief technology officer, Mira Murati, is a woman (and was CEO during last year's drama when Sam Altman was fired), only 122 of the company's 686 staffers whose jobs involve building AI systems are women. The disparity was even worse last year, when Glass.ai conducted a similar survey of LinkedIn profiles and found that women made up just 12 percent of OpenAI's research staff.

GENDER BIAS INCLUDE IN AI

When women and ethnic minorities are not playing a role in building important technologies like AI, it is unlikely anyone will uncover potential bias before it is built into the system.

Amazon, for example, once used a recruiting algorithm that filtered out women's resumes because they had been trained to send mostly men's resumes. This might not be the case if more women worked in the system at Amazon. 

Algorithms that scan chest X-rays have been shown to systematically under-diagnose female patients, according to the book Invisible Women written by Caroline Criado-Perez, again because so many medical AI models are trained on data that does not suit men.

INDUSTRIAL NORMAL

None of this will come as a surprise to female AI researchers familiar with the field's legacy of objectifying women. Even before the recent generative AI boom, academics were known to test the performance of their models by using them to apply makeup to women, or by replacing their jeans with miniskirts, according to Sasha Luccioni, a researcher at open source AI company Hugging Face. 

Every time he talks about this method, Luccioni says he faces resistance. “However, these are just benchmarks,” he wrote, pointing out that in academia, women are vastly underrepresented, at 12 percent of machine learning researchers.  

This is a complex problem that will take years to resolve because its roots lie in the education system and systemic cultural norms. But OpenAI and its peers could hinder many modern efforts to level the playing field, such as bringing more girls and women into science, technology, engineering, and mathematics (STEM) industries, if their systems continue to perpetuate stereotypes.

Eliminating bias in the data they use to train algorithms is one step to fixing the problem. Another way is to hire more female researchers to address this chronic imbalance. Expect their future products to be more interesting - and even dangerous - otherwise.

The latest generative AI tools are just as terrifying. Image generators have made women appear more sexualized than men, while an investigative report by Bloomberg News found that Stable Diffusion, an open-source AI image generator, tends to forget about the existence of women. It produces three times as many images of men as women. Men dominate images of high-paying jobs, such as engineers, CEOs or doctors, while women are depicted in lower-paying jobs such as housekeepers and cashiers.

Share this post