At least ten books by women about artificial intelligence (AI), both for and against
The AI Mirror | Shannon Vallor | 2024 | Casual
Unmasking AI | Joy Buolamwini | 2023 | Casual
Companies that claim to fear existential risk from AI could show a genuine commitment to safeguarding humanity by not releasing the AI tools they claim could end humanity.
Empire of AI | Karen Hao | 2025 | Casual
In 2023, Stanford researchers would create a transparency tracker to score AI companies on whether they revealed even basic information about their large deep learning models, such as how many parameters they had, what data they were trained on, and whether there had been any independent verification of their capabilities. All ten of the companies they evaluated in the first year, including OpenAI, Google, and Anthropic, received an F; the highest score was 54 percent.
The Algorithm | Hilke Schellmann | 2024 | Casual
I would read a text in German, my native language. After every question Christine asked, I read in German the Wikipedia entry for psychometrics, which deals broadly with measurements in psychology.
Here is what I read: <i>Die Psychometrie ist das Gebiet der Psychologie, das sich allgemein mit Theorie und Methode des psychologischen Messens befasst...</i> And so on and so forth. No words in English crossed my lips.
I thought after answering all the questions in German I would get an error message from the system saying it couldn't compute any scores.
I was surprised when I got a message with the results. In fact, the AI gave me a score of 6 out of 9 for English competency, and overall my skill level in English was deemed "competent."
Against Reduction | Noelani Arista, et al. | 2021 | Academic
A typical chilling forecast of AI is that it will be smarter, stronger, and more powerful than us, but the real fear should be that it might not be better. It could be instilled with values from our past, with less nuance, more bias, and replete with reductionist tropes.
Feminist AI | Jude Browne, et al. | 2024 | Academic
...such algorithms reinforce the status quo: those who have the most resources and the highest likelihood of success receive more resources. Through predictive algorithms, the past is recursively projected into the future, thus foreclosing options that could lead to more equitable distribution of resources and more diversity in the pool of those likely to succeed.
AI Needs You | Verity Harding | 2024 | Casual
As Jessica Montgomery of the University of Cambridge has highlighted, a survey from 2017 indicated that the public felt AI-enabled art was the least useful AI technology, and yet we've seen hundreds of millions of dollars invested into programs that use AI to generate images.
Future Tense | Martha Brockenbrough | 2024 | Casual
A computer can multiply two 20-digit numbers with speed and accuracy. This would be hard for even a math champ.
But a human can button a shirt, tie shoes, or fix a bowl of cereal with milk - things that would be tough to accomplish for an AI-powered robot.
Human beings think the multiplication challenge is hard. We think getting dressed and eating breakfast is easy. But when it comes to the computational power each task takes, we have it backward.
The AI Con | Emily M. Bender | 2025 | Casual
Corrado himself admits that he wouldn't want the tool to be a part of his family's "healthcare journey." But in the same breath, he says the large language model will take "the places in healthcare where AI can be beneficial and [expand] them by 10-fold." (We note that a tenfold increase on zero is still zero. So his statement might actually be technically true.)
Code-Dependent | Madhumita Murgia | 2024 | Casual
Empathy teaches us that everyone is flawed, yet still worthy of mercy. A risk score says the opposite: this is your digitally fixed reality, you have criminality inside you waiting to burst out. Your circumstances mean you don't deserve forgiveness.
Supremacy | Parmy Olson | 2024 | Casual
Imagine if a large food manufacturer like Unilever made increasingly delicious snacks but refused to put the ingredients on its packaging or explain how that food was made. That's essentially what OpenAI was doing. You could learn more about what was in a pack of Doritos than you could about a large language model.
The Atlas of AI | Kate Crawford | 2020 | Academic
There were categories for apples and airplanes, scuba divers and sumo wrestlers. But there were cruel, offensive, and racist labels, too: photographs of people were classified into categories like "alcoholic," "ape-man," "crazy," "hooker," and "slant eye." All of these terms were imported from WordNet's lexical database and given to crowdworkers to pair with images.
Robot Souls | Eve Poole | 2023 | Academic
But I think we would all want to argue that there is still something qualitatively different between AI learning to appreciate the colour red, and a human spontaneously doing so. In French this would be the difference between the verbs for knowing, savoir and connaître. Savoir is the kind of knowing that we can give AI; connaître, that familiarity with red, comes from somewhere else.
The New Age of Sexism | Laura Bates | 2025 | Casual
When users become desensitized to venting their frustration at Siri or Alexa for offering substandard responses or for not being clever or efficient enough, there is a risk that both they and others present, such as children growing up in homes where AI assistants are regularly used, absorb the belief that it is normal and acceptable to speak to women in a similar way.
Confronting Dystopia | Eva Paus | 2018 | Academic
Prophecies about the devastating impact of new technologies on jobs and working conditions are not new, going back to at least the early nineteenth century, when the Luddites smashed the steam-powered looms that were threatening their jobs.
Your Face Belongs to Us | Kashmir Hill | 2023 | Casual
When the prototype "automated people meter" successfully detected a person sitting on the couch, their name popped up above their head in a sans-serif white font on Turk's desktop computer. It was working perfectly until the film crew surprised Turk by bringing in a black Labrador.
The system tagged the dog as "Stanzi," the one woman in the experiment.
The Mind's Mirror | Daniela Rus | 2024 | Casual
Imagine a future AI that manages content recommendations on a social media site and is tasked with maximizing user engagement. With self-improvement capabilities, the AI system would enhance its models to better understand user behavior, preferences, and triggers. Over time, it might learn that sensational, polarizing, or even false content tends to keep users engaged longer than balanced, fact-based content. To fulfill its objective of maximizing engagement, the AI system would then preferentially serve more of this sensational content to users. If left unchecked, this recursively improving AI system would amplify misinformation, deepen societal divisions, even spark real-world conflicts.