That is right this moment’s version of The Obtain, our weekday e-newsletter that gives a day by day dose of what’s occurring on the planet of expertise.
5 methods criminals are utilizing AI
Synthetic intelligence has introduced an enormous enhance in productiveness—to the felony underworld.
Generative AI offers a brand new, highly effective device package that permits malicious actors to work way more effectively and internationally than ever earlier than. Over the previous yr, cybercriminals have largely stopped growing their very own AI fashions. As a substitute, they’re choosing tips with present instruments that work reliably.
That’s as a result of criminals need a simple life and fast features. For any new expertise to be well worth the unknown dangers related to adopting it—for instance, a better danger of getting caught—it must be higher and produce larger rewards than what they’re at present utilizing. Melissa Heikkilä, our senior AI reporter, has rounded up 5 methods criminals are utilizing AI now.
OpenAI’s newest blunder reveals the challenges dealing with Chinese language AI fashions
Final week’s launch of GPT-4o, a brand new AI “omnimodel”, was speculated to be an enormous second for OpenAI. However simply days later, it feels as if the corporate is in huge bother. From the resignation of most of its security staff to Scarlett Johansson’s accusation that it replicated her voice for the mannequin in opposition to her consent, it’s now in damage-control mode.
On prime of that, the info it used to coach GPT-4o’s tokenizer—a device that helps the mannequin parse and course of textual content extra effectively—is polluted by Chinese language spam web sites. In consequence, the mannequin’s Chinese language token library is filled with phrases associated to pornography and playing. This might worsen some issues which might be widespread with AI fashions: hallucinations, poor efficiency, and misuse.
However OpenAI is just not the one firm battling this downside: there are some steep challenges related to coaching giant language fashions to talk Chinese language. Learn our story to study extra.
—Zeyi Yang
This story is from China Report, our weekly e-newsletter providing you with the within monitor on tech in China. Join to obtain it in your inbox each Tuesday.
The must-reads
I’ve combed the web to search out you right this moment’s most enjoyable/vital/scary/fascinating tales about expertise.
1 AI simply bought rather less mysterious
Anthropic has delved into how synthetic neural networks work. (NYT $)
+ Understanding extra about how AI makes choices may assist us management it. (Wired $)
+ Massive language fashions can do jaw-dropping issues. However no person is aware of precisely why. (MIT Know-how Evaluation)
2 Google is testing adverts in its AI-generated search outcomes
Sponsored question solutions? No thanks. (Reuters)
+ Why you shouldn’t belief AI serps. (MIT Know-how Evaluation)
three China has created a chatbot skilled on the ideas of Xi Jinping
However we’ll have to attend to see how in style that’ll be, because it’s nonetheless a means off from being launched to the broader public. (FT $)
+ Why the Chinese language authorities is sparing AI from harsh laws—for now. (MIT Know-how Evaluation)
four Our ingesting water is main hacking goal
Default passwords are responsible. (IEEE Spectrum)
5 Humane is in search of a purchaser
Simply weeks after its AI pin gadget bought slated in opinions. (Bloomberg $)
6 How an enormous company coated up the risks of eternally chemical compounds
And stored promoting them afterwards. (New Yorker $)
+ The race to destroy PFAS, the eternally chemical compounds. (MIT Know-how Evaluation)
7 Contained in the struggle for America’s broadband
Marketing campaign teams are clashing with service suppliers over entry. (Ars Technica)
eight Sailboats are making a comeback
And the sails have had a high-tech makeover. (Economist $)
9 Can beef ever actually be climate-friendly?
The US branded a meat packer environmentally pleasant. Strain teams aren’t so certain. (Undark Journal)
+ How I discovered to cease worrying and love pretend meat. (MIT Know-how Evaluation)
10 Admire the great thing about Earth from the ISS
These new images are actually breathtaking. (The Atlantic $)
Quote of the day
“I want we had referred to as it ‘completely different intelligence’. As a result of I’ve my intelligence. I don’t want any synthetic intelligence.”
—Satya Nadella, Microsoft’s chief govt, is anxious about folks giving AI techniques an excessive amount of credit score, he tells Bloomberg.
The massive story
Bringing the lofty concepts of pure math all the way down to earth
April 2023
—Pradeep Niroula
Arithmetic has lengthy been offered as a sanctuary from confusion and doubt, a spot to go in the hunt for solutions. Maybe a part of the mystique comes from the truth that biographies of mathematicians usually paint them as otherworldly savants.
As a graduate pupil in physics, I’ve seen the work that goes into conducting delicate experiments, however the day by day grind of mathematical discovery is a ritual altogether overseas to me. And this sense is just bolstered by in style books on math, which frequently take the tone of a pastor shelling out sermons to the trustworthy.
Fortunately, there are methods to carry it again all the way down to earth. Well-liked math books search a more energizing tackle these outdated concepts, be it by baking recipes or hot-button political points. My verdict: Why not? It’s value a shot. Learn the total story.
We will nonetheless have good issues
A spot for consolation, enjoyable and distraction to brighten up your day. (Obtained any concepts? Drop me a line or tweet ’em at me.)
+ Spare a thought for the Kyles of the world, 706 of whom traveled to the town of Kyle, Texas, solely to be instructed they hadn’t damaged a world report.
+ Why are spirographs so hypnotic?
+ For those who’re into Impressionism, there’s an entire host of impressive-looking reveals happening this yr.
+ Right here’s what went down when the Seashore Boys met the Beatles.
