This Week’s Superior Tech Tales From Across the Internet (By way of April 1)


ChatGPT Is About to Revolutionize the Financial system. We Have to Determine What That Appears to be like Like.
David Rotman | MIT Know-how Evaluation
“The optimistic view: it can show to be a strong device for a lot of staff, enhancing their capabilities and experience, whereas offering a lift to the general economic system. The pessimistic one: corporations will merely use it to destroy what as soon as regarded like automation-proof jobs, well-paying ones that require artistic expertise and logical reasoning; a number of high-tech corporations and tech elites will get even richer, however it can do little for general financial progress.”

Why Exams Supposed for People May Not Be Good Benchmarks for LLMs Like GPT-4
Ben Dickson | VentureBeat
“In accordance with a technical report launched by OpenAI, GPT-4 performs impressively on bar exams, SAT math exams, and studying and writing exams. Nevertheless, exams designed for people is probably not good benchmarks for measuring LLMs’ capabilities. Language fashions embody data in intricate methods, typically producing outcomes that match or exceed common human efficiency. Nevertheless, the way in which they get hold of the data and use it’s usually incompatible with that of people. That may lead us to attract mistaken conclusions from check outcomes.”

The Unbelievable Zombie Comeback of Analog Computing
Charles Platt | Wired
“When outdated tech dies, it often stays useless. Nobody expects rotary telephones or including machines to come back crawling again from oblivion. Floppy diskettes, VHS tapes, cathode-ray tubes—they shall relaxation in peace. Likewise, we received’t see outdated analog computer systems in information facilities anytime quickly. They had been monstrous beasts: troublesome to program, costly to keep up, and restricted in accuracy. Or so I assumed. Then I got here throughout this confounding assertion: Bringing again analog computer systems in rather more superior types than their historic ancestors will change the world of computing drastically and ceaselessly. Significantly?”

The Woolly-Mammoth Meatball Is an All-Time Nice Meals Stunt
Yasmin Tayag | The Atlantic
i‘Sometimes sudden, humorous, or edgy, stunt meals are ‘pure advertising and marketing,’ Mark Lang, a advertising and marketing professor on the College of Tampa, instructed me. They work as a result of they’re bonkers sufficient to interrupt by the noise of social media and get folks speaking, he mentioned. However to this point, they’ve caught our consideration by twisting acquainted objects. Lab-grown meat, and all of the permutations of protein it makes attainable, is pushing us into a brand new period of stunt advertising and marketing, one involving meals folks might have by no means tried.”

A Massive Rover Goals to Be Like ‘UPS for the Moon’
Kenneth Chang | The New York Occasions
“[Founder and CEO Jaret Matthews] mentioned Astrolab would generate income by lifting and deploying cargo for patrons on the lunar floor. That might embody scientific devices. Sooner or later, the rover may assist construct lunar infrastructure. ‘Basically offering what I prefer to name last-mile mobility on the moon,’ Mr. Matthews mentioned. ‘You may sort of consider it like being UPS for the moon. And on this analogy, Starship is the container ship crossing the ocean, and we’re the native distribution answer.’i

Would Constructing a Dyson Sphere Be Value It? We Ran the Numbers.
Paul Sutter | Ars Technica
“What if we determined to construct a Dyson sphere round our solar? Might we do it? How a lot power wouldn’t it value us to rearrange our photo voltaic system, and the way lengthy wouldn’t it take to get our funding again? Earlier than we put an excessive amount of thought into whether or not humanity is able to this superb feat, even theoretically, we should always determine if it’s well worth the effort. Can we really obtain a web achieve in power by constructing a Dyson sphere?”

Cops Used Creepy Clearview AI a Million Occasions, CEO Says
Mack DeGeurin | Gizmodo
“Clearview AI, the shady US facial recognition agency whose surveillance tech is utilized by at the least 2,400 regulation enforcement businesses, says police have run practically 1,000,000 searches utilizing its service. The corporate’s database of photographs scraped from social media websites now reportedly numbers round 30 billion, a staggering 50% enhance from figures reported simply final yr. Regardless of repeated fines and years of pushback from civil liberties organizations, the figures counsel enterprise remains to be booming for Clearview.”


Why the AI Trade Might Stand to Sluggish Down a Little
Casey Newton | The Verge
“Tech protection tends to concentrate on innovation and the instant disruptions that stem from it. It’s sometimes much less adept at considering by how new applied sciences would possibly trigger society-level change. And but the potential for AI to dramatically have an effect on the job market, the knowledge setting, cybersecurity, and geopolitics—to call simply 4 issues—ought to provides us all cause to suppose larger.”

That Was Quick! Microsoft Slips Adverts Into AI-Powered Bing Chat
Devin Coldewey | TechCrunch
“Whereas nobody expects Microsoft, or Google, Amazon, Meta and all of the others to function these costly and computation-hungry language fashions out of the goodness of their hearts (assuming they’ve hearts and there’s good in them), it might be good to see slightly extra thought put into how promoting can higher be built-in. When the entire mannequin is altering, the apparent answer—that occurs to be rather a lot like one you used within the outdated instances—is unlikely to be the perfect.”

ChatGPT Can Substitute the Underpaid Staff Who Prepare AI, Researchers Say
Chloe Xiang | Motherboard
“In a brand new paper, political science researchers from the College of Zurich discovered that ChatGPT may outperform crowd-workers who carry out textual content annotation duties—that’s, labeling textual content for use in coaching an AI system. They discovered that ChatGPT may label textual content with extra accuracy and consistency than human annotators that they discovered on Mechanical Turk, an Amazon-owned crowdsourcing platform, in addition to skilled annotators resembling analysis assistants.”

Picture Credit score: Danist Soh / Unsplash