This is right now’s version of The Download, our weekday publication that gives a day by day dose of what’s occurring on the planet of expertise.
AI’s recursive self-improvement won’t come so shortly in spite of everything
The AI business’s boldest promise proper now could be that AI will quickly enhance itself, with virtually no want for human oversight. But a brand new research suggests it would take some time to get there.
Researchers discovered that AI brokers nonetheless can’t conduct open-ended AI analysis—free-form investigations with no clear-cut solutions that require the judgment and creativity wanted to make real breakthroughs.
The large query now could be how essential open-ended analysis is to recursive self-improvement—and whether or not AI methods can grind their method there with out it, just by bettering on narrower duties.
Find out why the outcomes could mood claims that recursive self-improvement is on the horizon.
—Michelle Kim
MIT Technology Review Narrated: what’s behind this summer time’s warmth, and why 2027 could possibly be worse
This summer time has been a scorcher for a lot of the Northern Hemisphere. June and July marked the most popular two-month stretch in Europe since record-keeping started, the contiguous US endured its hottest month on file in July, and South Korea noticed its highest-ever recorded temperature.
Climate change makes warmth waves extra possible and extra intense. But there’s one other issue at play: El Niño, which is already ramping up and is anticipated to have an even bigger impact on world temperatures subsequent 12 months.
Here’s what’s behind this summer time’s excessive warmth—and why 2027 could possibly be even hotter.
This is our newest story to be became an MIT Technology Review Narrated podcast, which we publish every week on Spotify and Apple Podcasts. Just navigate to MIT Technology Review Narrated on both platform, and comply with us to get all our new content material because it’s launched.
The must-reads
I’ve combed the web to search out you right now’s most enjoyable/necessary/scary/fascinating tales about expertise.
1 OpenAI has paused some mannequin work over security issues
It says its Astra mannequin reached a “important” threat threshold. (Guardian)
+ The slowdown units it other than Anthropic’s method. (Axios)
+ OpenAI has made safety updates after the Hugging Face hack. (Verge)
+ It’s additionally launched a model of ChatGPT for youngsters. (AP News)
2 Chinese humanoid maker Unitree surged 629% in its inventory market debut
The IPO has underscored China’s lead in humanoid robots. (Bloomberg $)
+ Unitree has develop into a key participant within the Sino-US tech conflict. (Reuters $)
+ It’s the world’s largest humanoid agency and is already worthwhile. (BBC)
+ There’s hidden human labor behind humanoids. (MIT Technology Review)
3 China is permitting Nvidia’s H200 chips into the mainland
ByteDance and Tencent every just lately acquired about 10,000 processors. (FT $)
+ A brand new Chinese AI mannequin might assist defenders and hackers alike. (Wired $)
4 5 Flock’s new AI device lets police determine drivers by their actions
It goes a lot additional than monitoring license plates. (Wired $)
+ Here’s what Flock’s defenders are lacking. (MIT Technology Review)
6 Prosecutors argue Meta intentionally hooked youngsters on Facebook and Instagram
The declare was made on the primary day of a landmark trial. (BBC)
+ A scientist is constructing a lacking map of childhood. (MIT Technology Review)
7 Even ICE has moral points with Meta’s sensible glasses
It’s banned them from the office over privateness issues. (NYT $)
+ Anduril and Meta are constructing sensible glasses for warfare. (MIT Technology Review)
8 X’s algorithm learns what you hate—and exhibits you extra of it
A research discovered the influence was even stronger amongst Democrats. (404 Media)
9 A physicist proposes that the universe could repeat itself without end
The identical galaxies, lives and occasions might recur endlessly. (New Scientist $)
10 A blind Egyptian developer has constructed an AI app that helps others “see”
It makes use of cameras to reply questions on customers’ environment. (Reuters $)
Quote of the day
“Arizona is Arizone. Illinois begins with a V. I imply, it’s loopy.”
—Stacey Morris, a Kentucky mother whose son introduced dwelling a packet of AI-generated instructional supplies, tells native station WDRB about a few of the errors.
One More Thing
The astonishing embryo fashions of Jacob Hanna
When the Palestinian stem-cell scientist Jacob Hanna was stopped whereas coming into the US final May, he knew the routine. Anything to declare? Any organic samples? But this time the brokers’ questions touched on a selected new subject: embryos.
Hanna didn’t have any specimens, but when he had, it will have been surprisingly arduous to say what they have been. That’s as a result of he makes a speciality of creating artificial embryo fashions, constructions that resemble actual embryos however don’t contain sperm, eggs, or fertilization.
They might present unprecedented views of the earliest levels of human improvement, whereas additionally creating a brand new method to produce cells for transplant drugs. But as these fashions develop into extra practical, they’re elevating tough questions on how far the science ought to go.
Take a better have a look at Hanna’s astounding embryo fashions—and the place they might lead.
— Antonio Regalado
We can nonetheless have good issues
A spot for consolation, enjoyable, and distraction to brighten up your day. (Got any concepts? Drop me a line.)
+ Take a guided digital tour by means of Spain’s astonishing cave work.
+ Explore virtually 2,000 prehistoric giants with the web’s largest dinosaur database.
+ The humble comfort retailer has been lovingly reimagined as a floating public art work.
+ If you’re trying to replicate on life or create a private time capsule, you may write an e-mail to your future self at FutureMe.