.png)
1 week ago
4
This is today's variation of The Download, our weekday newsletter that provides a regular dose of what's going connected successful the satellite of technology.
This information acceptable helps researchers spot harmful stereotypes successful LLMs
What’s new? AI models are riddled with culturally circumstantial biases. A caller information set, called SHADES, is designed to assistance developers combat the occupation by spotting harmful stereotypes and different kinds of favoritism that look successful AI chatbot responses crossed a wide scope of languages.
Why it matters: Although tools that spot stereotypes successful AI models already exist, the immense bulk of them enactment lone connected models trained successful English. They place stereotypes successful models trained successful different languages by relying connected instrumentality translations from English, which tin neglect to admit stereotypes recovered lone wrong definite non-English languages. To get astir these problematic generalizations, SHADES was built utilizing 16 languages from 37 geopolitical regions. Read the afloat story.
—Rhiannon Williams
MIT Technology Review Narrated: The 2nd question of AI coding is here
A drawstring of startups are racing to physique models that tin nutrient amended and amended software. They assertion it’s the shortest way to AGI.
This is our latest story to beryllium turned into a MIT Technology Review Narrated podcast, which we’re publishing each week connected Spotify and Apple Podcasts. Just navigate to MIT Technology Review Narrated connected either platform, and travel america to get each our caller contented arsenic it’s released.
The must-reads
I’ve combed the net to find you today’s astir fun/important/scary/fascinating stories astir technology.
1 Meta has launched its standalone AI app to rival ChatGPT
The Meta AI app combines its AI adjunct with a societal media feed. (The Verge)
+ It’s chiefly designed astir dependable conversations. (Bloomberg $)
+ Targeted ads are definite to follow. (TechCrunch)
2 Amazon won’t show tariff-induced terms rises aft all
Jeff Bezos rapidly sought to reassure Donald Trump it wasn’t happening. (WSJ $)
+ Big Tech’s marketplace worth has plummeted since Trump’s inauguration. (Economist $)
+ Tech leaders’ fealty to Trump is not being repaid successful kind. (Fast Company $)
3 OpenAI has rolled backmost an update that made ChatGPT ace chatty
Users complained it had abruptly go excessively sycophantic. (Ars Technica)
+ Sam Altman acknowledged the problem. (Bloomberg $)
4 Huawei is rushing to fulfil spot orders from Chinese clients
Now Nvidia is nary longer available, Huawei is blessed to measurement up. (FT $)
+ The UK’s semiconductor manufacture is softly bouncing back. (The Conversation)
5 The Gates Foundation is nether threat
The instauration is struggling with the Trump administration’s monolithic cuts to overseas aid. (NYT $)
6 We’re surviving successful a caller epoch of deepfake fraud
Fraudsters are manipulating video calls successful existent time. (404 Media)
+ An AI startup made a hyperrealistic deepfake of maine that’s truthful bully it’s scary. (MIT Technology Review)
7 What happens erstwhile we pain everlastingly chemicals?
Citizens successful Connecticut are paying the price. (Undark)
+ The contention to destruct PFAS, the everlastingly chemicals. (MIT Technology Review)
8 The fig of integer creators successful the US has exploded
They’re the fastest-growing assemblage of the country’s internet-dependent jobs. (Axios)
9 Why ChatGPT sounds truthful American
A caller survey sheds airy connected wherefore the chatbot lacks linguistic nuance. (Fast Company $)
10 The viral crystal bucket situation is back
More than a decennary aft it archetypal swept the internet. (WP $)
Quote of the day
“I’m not funny successful speechmaking thing that cipher said.”
—Emily M Bender, a computational-linguistics prof astatine the University of Washington, tells the Atlantic wherefore she refuses to usage AI substance generators.
One much thing
How adjacent are we to genuine “mind reading?”
Technically speaking, neuroscientists person been capable to work your caput for decades. It’s not easy, caput you. First, you indispensable prevarication motionless wrong a fMRI scanner, possibly for hours, portion you ticker films oregon perceive to audiobooks.
If you bash elite to endure claustrophobic hours successful the scanner, the bundle volition larn to make a bespoke reconstruction of what you were seeing oregon listening to, conscionable by analyzing however humor moves done your brain.
More recently, researchers person deployed generative AI tools, similar Stable Diffusion and GPT, to make acold much realistic, if not wholly accurate, reconstructions of films and podcasts based connected neural activity. So however adjacent are we to genuine “mind reading?” Read the afloat story.
—Grace Huckins
We tin inactive person bully things
A spot for comfort, amusive and distraction to brighten up your day. (Got immoderate ideas? Drop maine a line oregon skeet 'em astatine me.)+ Finally, immoderate bully news—Valerie the miniature dachshund has been rescued aft spending 529 days successful the Australian wilderness.
+ If you’re a genitor oregon teacher who’s having to enactment up with changeless shouts of “six seven,” remainder assured you’re not alone.
+ How is The Legend of Zelda” Majora’s Mask a 4th of a period old!?
+ Meanwhile successful the UK, we’re turning our beforehand lawns into meadows.
English (US) ·