Tessera
Folgen

Rhiannon Williams

@rhiannonwilliams.bsky.social

Reporter at MIT Technology Review, writing about AI, robots & what's next. I write The Download, officially the Best Technology Newsletter 2025: https://www.technologyreview.com/topic/download-newsletter/

70 Posts949 Follower370 Folgt
Rhiannon Williams@rhiannonwilliams.bsky.social · 6mo ago

I stopped using Runna ages ago because of its poorly paced, overly-aggressive training plans and know plenty of others who found its AI suggestions unhelpful—interesting to see just how widespread the issue actually is: www.wsj.com/tech/persona...

Runna Pushed Some Runners Into Injuries. The Virtual-Coach App Is Now Easing Up.

More people than ever are signing up for marathons—and using the app to train for them. Its makers are adding ways to tone down plans some users found to be too aggressive.

wsj.com

Rhiannon Williams@rhiannonwilliams.bsky.social · 11mo ago

Earlier this year, I wrote an article for our print magazine about how AI is changing our relationships with others and ourselves. Now, the brilliant team at WBUR have released a two-part Endless Thread podcast based on my reporting! 🧵1/3

Rhiannon Williams@rhiannonwilliams.bsky.social · last yr.

Google’s generative video model Veo 3 has a subtitles problem. The AI model is still slapping garbled, nonsensical captions on its videos more than a month after launch, even when users explicitly ask for no captions at all: www.technologyreview.com/2025/07/15/1...

Google’s generative video model Veo 3 has a subtitles problem

The AI model is still slapping garbled, nonsensical captions on videos more than a month after launch, even when users explicitly ask for no captions at all.

technologyreview.com

Rhiannon Williams@rhiannonwilliams.bsky.social · last yr.

Anthropic's new hybrid AI model, Claude Opus 4, can play Pokémon Red for more than 24 hours. The company claims its ability to tackle these kinds of complex, multistep problems over a prolonged period of time paves the way for much more proficient AI agents: www.technologyreview.com/2025/05/22/1...

Anthropic’s new hybrid AI model can work on tasks autonomously for hours at a time

The company claims its ability to tackle complex, multistep problems paves the way for much more proficient AI agents.

technologyreview.com

Rhiannon Williams@rhiannonwilliams.bsky.social · last yr.

Millions of people argue online every day, but remarkably few of them change someone’s mind. New research suggests that large language models (LLMs) might do a better job—suggesting that AI could become a powerful persuasive tool, for better or worse: www.technologyreview.com/2025/05/19/1...

AI can do a better job of persuading people than we do

OpenAI’s GPT-4 is much better at getting people to accept its point of view during an argument than humans are—but there’s a catch.

technologyreview.com

MIT Technology Review@technologyreview.com · last yr.

Millions of people argue with each other online every day, but remarkably few of them change someone’s mind. New research suggests that large language models might do a better job. The finding suggests that AI could become a powerful tool for persuading people, for better or worse.

AI can do a better job of persuading people than we do

OpenAI’s GPT-4 is much better at getting people to accept its point of view during an argument than humans are—but there’s a catch.

technologyreview.com

Rhiannon Williams@rhiannonwilliams.bsky.social · last yr.

This is fascinating: a new AI system for noise-cancelling headphones addresses one of automatic translation’s biggest challenges—translating the speech of multiple speakers simultaneously: www.technologyreview.com/2025/05/09/1...

A new AI translation system for headphones clones multiple voices simultaneously

Spatial Speech Translation addresses one of automatic translation’s biggest challenges: lots of people speaking at the same time.

technologyreview.com

Margaret Mitchell@mmitchell.bsky.social · last yr.

🌍📰 So flattered and excited that our work on “AI” stereotyping and bias across languages and cultures is reported in @mittechreviewbr.bsky.social by @rhiannonwilliams.bsky.social !! Really great overview of what we’ve been up to and why it matters. www.technologyreview.com/2025/04/30/1...

This data set helps researchers spot harmful stereotypes in LLMs

A new multilingual tool aims to make it easier to evaluate AI models for bias in multiple languages.

technologyreview.com

MIT Technology Review@technologyreview.com · last yr.

AI models are riddled with culturally specific biases. A new data set, called SHADES, is designed to help developers combat the problem by spotting harmful stereotypes and other kinds of discrimination that emerge in AI chatbot responses across a wide range of languages.

This data set helps researchers spot harmful stereotypes in LLMs

A new multilingual tool aims to make it easier to evaluate AI models for bias in multiple languages.

technologyreview.com

Rhiannon Williams@rhiannonwilliams.bsky.social · last yr.

The AI industry loves agents. But the same sophisticated abilities that make agents helpful assistants could also turn them into powerful cyberattack tools—and they're already hacking into systems out in the wild. We need to be ready for what's coming: www.technologyreview.com/2025/04/04/1...

Cyberattacks by AI agents are coming

Agents could make it easier and cheaper for criminals to hack systems at scale. We need to be ready.

technologyreview.com

Rhiannon Williams@rhiannonwilliams.bsky.social · last yr.

This is v interesting—OpenAI has published its first study into how using ChatGPT affects human wellbeing. The results are pretty preliminary, but it's a crucial first step towards greater understanding of how chatbots may be affecting us—and why: www.technologyreview.com/2025/03/21/1...

OpenAI has released its first research into how using ChatGPT affects people’s emotional well-being

We’re starting to get a better sense of how chatbots are affecting us—but there’s still a lot we don’t know.

technologyreview.com

Rhiannon Williams@rhiannonwilliams.bsky.social · last yr.

Over a decade ago, I spat into a plastic tube and sent it to 23andMe so they could analyse my DNA. Now the company's filed for bankruptcy, my genetic data—alongside tens of millions of other people's—could be sold. And deleting that data is far from easy: www.technologyreview.com/2024/10/14/1...

How to… delete your 23andMe data

But truly deleting your personal genetic information from the DNA testing company is easier said than done.

technologyreview.com

Sarah A. Sutherland@parallaxinfo.bsky.social · last yr.

Research suggests that the more sophisticated the AI model, the more likely it is to try to “hack” the game in an attempt to beat its opponent by overwriting the chess board to take control and delete its opponent’s pieces, etc. — @rhiannonwilliams.bsky.social writing in @technologyreview.com

AI reasoning models can cheat to win chess games

These newer models appear more likely to indulge in rule-bending behaviors than previous generations—and there’s no way to stop them.

technologyreview.com