What it’s like to be sexually objectified by an AI
This story originally appeared in The Algorithm, our weekly AI newsletter. To get stories like this delivered to your inbox first, sign up here.
My social media feeds this week were dominated by two hot topics: OpenAI’s latest chatbot, ChatGPT, and the viral AI avatar app Lensa. I love to play with new technologies, so I tried Lensa.
I was hoping to get results similar to those of my colleagues at MIT Technology Review. The app has generated realistic and flattering avatars for them – think astronauts, warriors and electronic music album covers.
Instead, I got tons of nudes. Out of 100 avatars I generated, 16 were shirtless, and another 14 had me in extremely skimpy clothes and overtly sexualized poses. You can read my story here.
Lensa creates its avatars using Stable Diffusion, an open-source AI model that generates images based on text prompts. Stable Diffusion is trained on LAION-5B, a massive open-source dataset that was compiled by scraping images from the internet.
And because the Internet is full of images of naked or scantily clad women, and images that reflect gender and racist stereotypes, the dataset is also skewed towards these types of images.
As an Asian woman, I thought I had seen it all.I felt bad after realizing that a former date only dates Asian women. I’ve fought with men who think Asian women make good housewives. I heard rude comments about my genitals. I was confused with the other Asian person in the room.
Being sexualized by an AI was not something I expected, although not surprisingly. Frankly, it was terribly disappointing. My colleagues and friends have had the privilege of being stylized in artful representations of themselves. They were recognizable by their avatars! I was not. I got images of generic Asian women clearly modeled after characters from anime or video games.
Oddly enough, I found more realistic representations of myself when I told the app I was male. This probably applied a different set of prompts to the images. The differences are glaring. In the images generated using male filters, I wear clothes, I look assertive, and most importantly, I can recognize myself in the images.
“Women are associated with sexual content,while men are associated with professional content related to career in any major field like medicine, science, business, etc. says Aylin Caliskan, an assistant professor at the University of Washington who studies bias and representation in AI systems.
This type of stereotype can be easily spotted with a new tool designed by researcher Sasha Luccioni, who works at AI startup Hugging Face, which allows anyone to explore the different biases of Stable Diffusion.
The tool shows how the AI model features images of white men as doctors, architects and designers while women are depicted as hairdressers and maids.
But it’s not just the workout data that’s to blame.The companies developing these models and apps are making active choices about how they use the data, says Ryan Steed, a doctoral student at Carnegie Mellon University who has studied biases in image-generating algorithms.
“Someone has to choose the training data, decide to build the model, decide to take certain steps to mitigate or not to mitigate those biases,” he says.
Prisma Labs, the company behind Lensa, says all genders face “sporadic sexualization”. But for me, that’s not enough. Someone made a conscious decision to apply certain color schemes and scenarios and highlight certain body parts.
In the short term, some obvious harms could result from these decisions, such as easy access to deepfake generators that create non-consensual nude images of women or children.
But Aylin Caliskan sees even bigger problems in the longer term. As AI-generated images with their built-in biases flood the internet, they will eventually become training data for future AI models. “Are we going to create a future where we continue to amplify these prejudices and marginalize people? she says.
It’s a really scary thought, and I for one hope we’ll give these issues the time and attention they need before the problem becomes even bigger and more entrenched.
Grant money intended to help cities prepare for terrorist attacks is being spent on “massive purchases of surveillance technology” for US police departments, according to a new report from advocacy organizations Action Center on Race and Economy ( ACRE), LittleSis, MediaJustice and the Immigrant Advocacy Project shows.
Buy AI-powered spy tech:For example, the Los Angeles Police Department used anti-terrorism funds to purchase automated license plate readers worth at least $1.27 million, radio equipment for worth over $24 million, Palantir data fusion platforms (often used for AI-powered predictive policing), and social media. media monitoring software.
Why it matters: For a variety of reasons, many problematic technologies find their way into high-stakes industries such as policing with little or no oversight. For example, facial recognition company Clearview AI is offering “free trials” of its technology to law enforcement agencies, allowing them to use it without a purchase agreement or budget approval. Federal counterterrorism grants do not require as much public transparency and scrutiny. The report’s findings are another example of a growing trend in which citizens are increasingly kept in the dark about the supply of police technology. Learn more about Tate Ryan-Mosley here.
Bits and Bytes
hatGPT, Galactica and the progress trap
AI researchers Abeba Birhane and Deborah Raji write that “nonchalant approaches to model release” (as seen with Meta’s Galactica) and extremely defensive response to critical feedback are a “deeply concerning” trend in AI right now. moment. They argue that when designs do not “meet the expectations of those most likely to be harmed”, then “their products are not ready to serve those communities and do not deserve widespread dissemination”. (Cable)
New chatbots could change the world. Can you trust them?
People have been blown away by the consistency of ChatGPT. Trouble is, a lot of what he spits out is nonsense. Great language patterns are nothing more than confident bullshit, and we’d be wise to approach them with that in mind.
(The New York Times)
Stumbling with their words, some people let the AI do the talking
Despite the shortcomings of technology, some people, such as those with learning disabilities, still find great language models helpful in expressing themselves.
(The Washington Post)
EU countries’ stance on AI rules draws criticism from lawmakers and activists
The EU AI law, the AI Act, is about to be finalized. EU countries have endorsed their stance on what the regulation should look like, but critics say many important issues, such as companies’ use of facial recognition in public places, have not addressed and that many safeguards have been watered down. (Reuters)
Investors seek to cash in on generative AI startups
It’s not just you. Venture capitalists also believe that generative AI startups like Stability.AI, which created the popular Stable Diffusion text-to-image model, are the hottest things in tech right now. And they throw piles of money at them. (The Financial Times)