1 year ago
95 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.
The UN conscionable handed retired an urgent clime to-do list. Here’s what it says.
Time is moving abbreviated to bounds planetary warming to 1.5°C (2.7 °F) supra preindustrial levels, but determination are feasible and effectual solutions connected the table, according to a caller UN clime report.
Despite decades of warnings from scientists, planetary greenhouse-gas emissions are inactive climbing, hitting a grounds precocious successful 2022. If humanity wants to bounds the worst effects of clime change, yearly greenhouse-gas emissions volition request to beryllium chopped by astir fractional betwixt present and 2030, according to the report.
That volition beryllium analyzable and expensive. But it is nevertheless doable, and the UN listed a fig of circumstantial ways we tin execute it. Read the afloat story.
—Casey Crownhart
How radical are utilizing GPT-4
Last week was aggravated for AI news, with a flood of large merchandise releases from a fig of starring companies. But 1 announcement outshined them all: OpenAI’s caller multimodal ample connection model, GPT-4. William Douglas Heaven, our elder AI editor, got an exclusive preview. Read astir his archetypal impressions.
Unlike OpenAI’s viral deed ChatGPT, which is freely accessible to the wide public, GPT-4 is presently accessible lone to developers. It’s inactive aboriginal days for the tech, and it’ll instrumentality a portion for it to provender done into caller products and services. Still, radical are already investigating its capabilities retired successful the open. Read about immoderate of the astir amusive and absorbing ways they’re doing that, from hustling up wealth to penning codification to reducing doctors’ workloads.
—Melissa Heikkilä
Melissa’s communicative is from The Algorithm, her play AI newsletter. Sign up to person it successful your inbox each Monday.
Language models mightiness beryllium capable to self-correct biases—if you inquire them
The news: Large connection models are infamous for spewing toxic biases. But if the models are ample enough, and humans person helped bid them, past they whitethorn beryllium capable to self-correct for immoderate of these biases, a caller insubstantial from AI laboratory Anthropic has found. Remarkably, each we person to bash is ask.
How they did it: The squad of researchers wanted to cognize if simply asking these models to nutrient output that was unbiased—without adjacent having to specify what they meant by bias—would beryllium capable to change what they produced. They recovered that conscionable prompting a exemplary to marque definite its answers didn’t trust connected stereotyping had a dramatically affirmative effect connected its output.
The significance: The enactment raises the evident question whether this “self-correction” could and should beryllium baked into connection models from the start. Read the afloat story.
—Niall Firth
The must-reads
I’ve combed the net to find you today’s astir fun/important/scary/fascinating stories astir technology.
1 We don’t cognize however to woody with the problems AI creates
Maybe we should beryllium pumping the brakes, not accelerating. (Vox)
+ How to halt worrying and larn to emotion your AI colleague. (WP $)
+ Generative AI is changing everything. But what’s near erstwhile the hype is gone? (MIT Technology Review)
2 China’s apical chipmakers person been granted caller powers
They’ll person tighter power implicit state-backed probe and greater entree to subsidies. (FT $)
+ Chinese chips volition support powering your mundane life. (MIT Technology Review)
3 A Meta manager was wiretapped by Greek authorities
Artemis Seaford, who is simply a US and Greek national, was spied connected for a year. (NYT $)
4 Amazon is readying to chopped different 9,000 jobs
Just months aft it laid disconnected much than 18,000 workers. (CNBC)
+ Amazon’s idiosyncratic national is facing a bid of setbacks. (NYT $)
5 The locations of US borderline surveillance towers are being made public
The Electronic Frontier Foundation has mapped adjacent to 300 towers on the US-Mexico border. (The Intercept)
+ How US constabulary usage counterterrorism wealth to bargain spy tech. (MIT Technology Review)
6 College coding classes aren’t ever what they seem
Some universities outsource bundle footwear camps to unregulated 3rd parties. (Wired $)
7 TikTok’s depressing algorithm loops tin beryllium pugnacious to break
There’s nary casual mode to accidental ‘please halt showing maine this.’(The Atlantic $)
+ The app has 150 cardinal monthly progressive users successful the US, now. (Reuters)
+ When my dada was sick, I started Googling grief. Then I couldn’t flight it. (MIT Technology Review)
8 It costs a batch much to complaint EVs connected the thoroughfare than astatine home
It’s besides cheaper to complaint overnight. (Reuters)
+ Ecuador’s taxi drivers privation EVs, but interest astir the deficiency of chargers. (Rest of World)
+ How does an EV artillery really work? (MIT Technology Review)
9 Do we privation to speech to chatbots, really?
Just due to the fact that we can, doesn’t mean we should. (Slate $)
+ A US legislator wants to cognize however chatbot makers volition support children. (Bloomberg $)
10 China wants its residents to find love
Ideally done its caller state-sponsored dating app, Palm Guixi. (The Guardian)
Quote of the day
“This is simply a headwind compared to the hurricane of the dotcom crash.”
—Manish Madhvani, managing spouse of exertion concern steadfast GP Bullhound, tells the Financial Times that comparisons betwixt today’s tech downturn and the dotcom bust are wildly overblown.
The large story
This idiosyncratic is trying to make an accessible, unhackable voting machine
November 2022
For the past 19 years, machine subject prof Juan Gilbert has immersed himself successful possibly the astir contentious statement implicit predetermination medication successful the United States—what role, if any, touch-screen ballot-marking devices should play successful the voting process.
While advocates assertion that physics voting systems tin beryllium comparatively secure, amended accessibility, and simplify voting and ballot tallying, critics person argued that they are insecure and should beryllium utilized arsenic infrequently arsenic possible.
As for Gilbert? He claims he’s yet invented “the astir unafraid voting exertion ever created.” And he’s invited respective of the astir respected and vocal critics of voting exertion to beryllium his point. Read the afloat story.
—Spencer Mestel
