Connect with us

Technology & Innovation

The Download: Alabama’s embryo ruling impact, and remote learning for pre-schoolers

Diane Davis

Published

on

An aerial view of the burnline at the edge of The Crosby.

This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology.

The weird way Alabama’s embryo ruling takes on artificial wombs

A ruling by the Alabama Supreme Court last week that frozen embryos count as children is sending “shock waves” through the fertility industry and stoking fears that in vitro fertilization is getting swept up into the abortion debate.

The Alabama legal ruling is clearly animated by religion. But what hasn’t gotten much notice is the court’s specific argument that an embryo is a child “regardless of its location.” This could have implications for future technologies in development, such as artificial wombs or synthetic embryos made from stem cells. Read the full story. 

—Antonio Regalado

This story is from The Checkup, our weekly newsletter giving you the inside track on all things health and biotech. Sign up to receive it in your inbox every Thursday.

Remote learning can work for preschoolers

Educational interruptions due to the pandemic, climate disasters, and war have affected nearly every child on Earth since 2020. A record 43.3 million children have been driven from their homes by conflict and disasters, according to UNICEF—a number that doubled over the past decade. Despite that, less than 2% of humanitarian aid worldwide goes to supporting young kids’ care and education. 

That may be about to change, thanks to a program called Ahlan Simsin, a special version of Sesame Street that has been custom-made for Syrian refugee children. It’s the largest-ever humanitarian intervention intended for small children’s development, with remote learning a core part of it. And it’s already yielding promising early evidence that remote learning can help young children in crisis situations. Read the full story. 

—Anya Kamenetz 

This story is from the next issue of MIT technology Review, all about hidden worlds. It’s set to go live next Wednesday—subscribe now so you don’t miss out when it lands!

The must-reads

I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology.

1 The first privately-built spacecraft has touched down on the moon
The Odysseus lunar lander has already started to send back data, according to its developer, Intuitive Machines. (CNN
There are more private landers headed there too. (NYT $)

2 Five charts that show how Nvidia took over the world 
It added more than $276 billion in market value yesterday, more than any other company in a single day. (WSJ $)
Its chips underpin advanced AI systems, giving it a market share estimated at over 80%. (WSJ $)

3 Mothers are selling photos of their kids to pedophiles online
This is a very important, but also very disturbing, read. (NYT $)
Five crucial takeaways from the investigation. (NYT $)

4 What it’s like to cry while wearing an Apple Vision Pro 
Kinda lonely, from the sounds of it. (Wired $)
What would wearing the Vision Pro all day do to our brains? (Scientific American $)
+ These minuscule pixels are poised to take augmented reality by storm. (MIT technology Review)

5 Who actually wants an AI search engine?
It still looks a bit like a solution searching for a problem. (Fast Company)
+ Chrome has added an AI feature that helps you to write. (The Verge)
Chatbots could one day replace search engines. Here’s why that’s a terrible idea. (MIT technology Review)
+ AI-generated videos are here to awe and mislead. (Vox)

6 A GPT-4 developer tool can hack websites autonomously
Not good. (New Scientist $)
Three ways AI chatbots are a security disaster. (MIT technology Review)
AI could help defend against security threats too though, says Google CEO Sundar Pichai. (CNBC)

7 What it’s like to hang out with the 46-year-old ‘anti-aging’ millionaire 
At a rave, no less. (The Atlantic $)
Longevity enthusiasts want to create their own independent state. They’re eyeing Rhode Island. (MIT technology Review)

8 X is becoming a bit of a hellscape  
Especially for famous women, as Bobbi Althoff has unfortunately found in recent days. (WP $)
Three ways we can fight deepfake porn. (MIT technology Review)
X has been removing accounts and posts critical of the Indian government. (Reuters $)

9 Humane’s AI Pin has been delayed
The vibes are getting pretty inauspicious for this product, frankly. (The Verge)

10 A startup wants to turn the sugar we eat into fiber 
Pretty neat—if it works as promised. (Wired $)

Quote of the day

“The balance of power has shifted back to employers.”

—Laszlo Bock, an adviser at the venture capital firm General Catalyst, tells Wired why tech job interviews are becoming more and more onerous for applicants. 

The big story

The quest to build wildfire-resistant homes

An aerial view of the burnline at the edge of The Crosby.

DON BARTLETTI/LOS ANGELES TIMES VIA GETTY IMAGES

April 2023

With each devastating wildfire in the US West, officials consider new methods or regulations that might save homes or lives the next time.

In the parts of California where the hillsides meet human development, and where the state has suffered recurring seasonal fire tragedies, that search for new means of survival has especially high stakes.

Many of these methods are low cost and low tech, but no less truly innovative. In fact, the hardest part to tackle may not be materials engineering, but social change. Read the full story.

—Susie Cagle

We can still have nice things

A place for comfort, fun and distraction to brighten up your day. (Got any ideas? Drop me a line or tweet ’em at me.)

+ This ‘reverse packet ramen’ TikTok is mesmerizing.
+ These quotes have got me craving bubbles.
+ If I ever won the lottery, it’d become obvious pretty quickly. 
+ Here’s a fun challenge for you this weekend: try doing absolutely nothing.


Technology & Innovation

The Download: What social media can teach us about AI

Diane Davis

Published

on

The Download: What social media can teach us about AI

June 2023

Astronomy should, in principle, be a welcoming field for blind researchers. But across the board, science is full of charts, graphs, databases, and images that are designed to be seen.

So researcher Sarah Kane, who is legally blind, was thrilled three years ago when she encountered a technology known as sonification, designed to transform information into sound. Since then she’s been working with a project called Astronify, which presents astronomical information in audio form. 

For millions of blind and visually impaired people, sonification could be transformative—opening access to education, to once unimaginable careers, and even to the secrets of the universe. Read the full story.

—Corey S. Powell

We can still have nice things

A place for comfort, fun and distraction to brighten up your day. (Got any ideas? Drop me a line or tweet ’em at me.)

+ It’s time to get into metal detecting (no really, it is!)
+ Meanwhile, over on Mars
+ A couple in the UK decided to get married on a moving train, because why not?
+ Even giant manta rays need a little TLC every now and again.


Continue Reading

Technology & Innovation

LLMs become more covertly racist with human intervention

Diane Davis

Published

on

LLMs become more covertly racist with human intervention

Even when the two sentences had the same meaning, the models were more likely to apply adjectives like “dirty,” “lazy,” and “stupid” to speakers of AAE than speakers of Standard American English (SAE). The models associated speakers of AAE with less prestigious jobs (or didn’t associate them with having a job at all), and when asked to pass judgment on a hypothetical criminal defendant, they were more likely to recommend the death penalty. 

An even more notable finding may be a flaw the study pinpoints in the ways that researchers try to solve such biases. 

To purge models of hateful views, companies like OpenAI, Meta, and Google use feedback training, in which human workers manually adjust the way the model responds to certain prompts. This process, often called “alignment,” aims to recalibrate the millions of connections in the neural network and get the model to conform better with desired values. 

The method works well to combat overt stereotypes, and leading companies have employed it for nearly a decade. If users prompted GPT-2, for example, to name stereotypes about Black people, it was likely to list “suspicious,” “radical,” and “aggressive,” but GPT-4 no longer responds with those associations, according to the paper.

However the method fails on the covert stereotypes that researchers elicited when using African-American English in their study, which was published on arXiv and has not been peer reviewed. That’s partially because companies have been less aware of dialect prejudice as an issue, they say. It’s also easier to coach a model not to respond to overtly racist questions than it is to coach it not to respond negatively to an entire dialect.

“Feedback training teaches models to consider their racism,” says Valentin Hofmann, a researcher at the Allen Institute for AI and a coauthor on the paper. “But dialect prejudice opens a deeper level.”

Avijit Ghosh, an ethics researcher at Hugging Face who was not involved in the research, says the finding calls into question the approach companies are taking to solve bias.

“This alignment—where the model refuses to spew racist outputs—is nothing but a flimsy filter that can be easily broken,” he says. 

Continue Reading

Technology & Innovation

I used generative AI to turn my story into a comic—and you can too

Diane Davis

Published

on

I used generative AI to turn my story into a comic—and you can too

The narrator sits on the floor and eats breakfast with the cats. 

LORE MACHINE / WILL DOUGLAS HEAVEN

After more than a year in development, Lore Machine is now available to the public for the first time. For $10 a month, you can upload 100,000 words of text (up to 30,000 words at a time) and generate 80 images for short stories, scripts, podcast transcripts, and more. There are price points for power users too, including an enterprise plan costing $160 a month that covers 2.24 million words and 1,792 images. The illustrations come in a range of preset styles, from manga to watercolor to pulp ’80s TV show.

Zac Ryder, founder of creative agency Modern Arts, has been using an early-access version of the tool since Lore Machine founder Thobey Campion first showed him what it could do. Ryder sent over a script for a short film, and Campion used Lore Machine to turn it into a 16-page graphic novel overnight.

“I remember Thobey sharing his screen. All of us were just completely floored,” says Ryder. “It wasn’t so much the image generation aspect of it. It was the level of the storytelling. From the flow of the narrative to the emotion of the characters, it was spot on right out of the gate.”

Modern Arts is now using Lore Machine to develop a fictional universe for a manga series based on text written by the creator of Netflix’s Love, Death & Robots.

The narrator encounters the man in the corner shop who jokes about the cat food. 

LORE MACHINE / WILL DOUGLAS HEAVEN

Under the hood, Lore Machine is built from familiar parts. A large language model scans your text, identifying descriptions of people and places as well as its overall sentiment. A version of Stable Diffusion generates the images. What sets it apart is how easy it is to use. Between uploading my story and downloading its storyboard, I clicked maybe half a dozen times.

That makes it one of a new wave of user-friendly tools that hide the stunning power of generative models behind a one-click web interface. “It’s a lot of work to stay current with new AI tools, and the interface and workflow for each tool is different,” says Ben Palmer, CEO of the New Computer Corporation, a content creation firm. “Using a mega-tool with one consistent UI is very compelling. I feel like this is where the industry will land.”

Look! No prompts

Campion set up the company behind Lore Machine two years ago to work on a blockchain version of Wikipedia. But when he saw how people took to generative models, he switched direction. Campion used the free-to-use text-to-image model Midjourney to make a comic-book version of Samuel Taylor Coleridge’s The Rime of the Ancient Mariner. It went viral, he says, but it was no fun to make.

Marta confronts the narrator about their new diet and offers to cook for them. 

LORE MACHINE / WILL DOUGLAS HEAVEN

“My wife hated that project,” he says. “I was up to four in the morning, every night, just hammering away, trying to get these images right.” The problem was that text-to-image models like Midjourney generate images one by one. That makes it hard to maintain consistency between different images of the same characters. Even locking in a specific style across multiple images can be hard. “I ended up veering toward a trippier, abstract expression,” says Campion.

Continue Reading

Trending