Dictate a text message into your smartphone and you’ll quickly realize that automatic speech recognition technology is far from perfect. But with recent improvements in software, many scientists are optimistic that real-time voice transcription with stenographer-level accuracy is just around the corner.

Last year, Microsoft created a speech recognition system that bested human transcriptionists on a standard set of recorded conversations scientists have been using as a benchmark for the last two decades.

“It wasn’t a real-time system,” Xuedong Huang, Microsoft’s chief speech scientist, told The Atlantic. “It was very much like we wanted to see, with all the horsepower we have, what is the limit. But the real-time system is not that far off.”

At the company’s annual build conference in May, Microsoft executive Harry Shum demonstrated a real-time transcription service that tied the spoken words of his presentation to individual PowerPoint slides. Real-time transcription isn’t always necessary for many transcription tasks, though. Journalists and other professionals who routinely transcribe recorded audio of interviews, meetings, and conference calls are turning to high-end, web-browser-based apps such as SwiftScribe—a product of Chinese internet company Baidu—that can transcribe up to an hour of recorded conversation with error rates of 5 percent or less, saving hours of tedious labor.

SwiftScribe developer Ryan Prenger and others believe improvements in automatic speech recognition will increase the supply and demand for transcripts, leading to a world in which almost every audio source—videos, radio shows, personal phone calls—becomes accessible as text.

“There could be a virtuous circle where more people expect more of their audio that they produce to be transcribed, because it’s now cheaper and easier to get things transcribed quickly,” Prenger told The Atlantic. “And so, it becomes the standard to transcribe everything.”

“Multi” elevator Handout

Going up? Down? Sideways?

Basic skyscraper design hasn’t changed in decades, largely because those designs have been constrained by the need for multiple elevator shafts. But in what some are calling the biggest innovation in elevator technology since its invention more than 160 years ago, German engineering firm ThyssenKrupp AG in June unveiled an 800-foot test tower with a full-scale version of “Multi,” its cable-free elevator system.

Running on rails and using motors like those on magnetic levitation trains, the Multi elevator can move both vertically and horizontally, with multiple cabins in a single shaft. Such a system should allow architects more flexibility in the design of large buildings.

The Multi elevator could cost up to five times more than a conventional system, according to Wired UK. But ThyssenKrupp claims floor space savings of as much as 40 percent will more than compensate. —M.C.

Hadrian X Handout

Brick by brick

U.S. construction equipment manufacturer Caterpillar has invested $2 million in Fastbrick Robotics, a company in Perth, Australia, that has developed a bricklaying robot.

The robot, called Hadrian X, works from a 3-D computer-aided design (CAD) model of a structure. From a single location, a 28-foot, truck-mounted boom moves the robotic bricklayer sequentially around the building site, laying hundreds of structural bricks (or blocks) per hour, complete with mortar—enough to construct around 150 houses per year.

Fastbrick Robotics will work with Caterpillar to commercialize the technology, collaborating on manufacturing, sales, and service, according to the website Australian Manufacturing. —M.C.