How firefighters help you get your pizza faster

When Thomas Melia, an engineer for the City of Salinas Fire Department in California, was dispatched earlier this year on a medical call to an assisted-living facility, his routing app told him where to park the truck. But when he pulled into the sprawling multi-unit complex, the ambulance service that had already arrived was parked near the front door, far from the apartment in need.

• “We looked and knew this unit is going to be right by the elevator by the back door,” Melia said.
• In the past, in order to make sure they parked in a spot where they could easily access a particular apartment and nearby standpipes, hydrants and sprinklers, the firefighters would have to flip through a “target hazard” binder full of printouts while frantically adjusting masks and tool belts en route to an emergency.
• Now, their fireproof mobile devices tell them not only the best route to take there, but the best spot to park the truck. That means Melia’s captain doesn’t have to flip through a binder as often.
• “It increases safety by having more eyes on the road,” said Melia. “You can start to come up with a plan before you even get there.”

Those firefighters rely on routing technology from Beans.ai, one of many companies building unique data sets and AI-based software that make the final steps to a destination speedier and more efficient.

• Founded over three years ago, Beans.ai was not originally intended strictly as a service for emergency responders.
• However, it turned out fire departments and government agencies were useful places to look when foraging for data that would help food and package delivery workers find the closest parking spot to a destination.
• Not only did the company temporarily borrow the Salinas Fire Department’s binders of maps to input data into its system, but Beans.ai even used freedom-of-information requests in the hopes of retrieving information such as physical maps that show building layouts, said Nitin Gupta, its co-founder and CEO.

Hunting for last-foot data could pay off as the rise of quick-commerce delivery services from companies such as Buyk, Gopuff and Gorillas spur competition and development in the delivery-routing tech arena. More-established companies including DoorDash have been gathering delivery logistics data to train models used in their delivery fulfillment apps for years.

• Last November, Coresight Research estimated that retail sales in the overall quick-commerce market would total $20 billion to $25 billion in the U.S. in 2021, around 10% of the research company’s share of estimated U.S. online consumer packaged goods sales for the year.
• New players like Fridge No More, JOKR and 1520 that promise 15-minute deliveries “have intensified instant needs in terms of speed promises,” according to Coresight.
• Companies in the quick-commerce sector will need accurate data and tools to get items to people at a rapid speed or they’ll risk losing out on customers who can easily switch to another service, said Christoph Herzig, head of Fleet Applications for Here Technologies, which sells its routing application for delivery drivers to businesses.
• “The value of location technology becomes even more important because it’s all about early user conversion,” Herzig said.

While helping to make emergency calls and deliveries more efficient, emerging technologies used for delivery routing and fleet management also create ethical questions related to worker privacy and civil rights.

• Apps used by workers that merge personal and professional device use are “undermining workers’ basic human right to disconnect” especially “when workers are required to use personal devices that deliver data to employers, which can be used against them,” wrote Wilneida Negrón in a 2021 report for Coworker.org.
• The report pointed to threats to worker rights that can be exacerbated by data collection and algorithmic technologies such as increased worker monitoring, wage theft and labor-organizing surveillance.
• When FedEx drivers or other delivery drivers use Beans.ai apps to assist in routing, they can turn off location tracking while still accessing static routing information, said Gupta, who added that Beans.ai does not sell its data as a separate product.
• Still, the company has added features and capabilities that some drivers might find invasive: It uses phone accelerator data to determine whether someone is driving, walking or idle, as well as camera footage data from a dashcam provider to help determine whether deliveries happened, if drivers parked where they say they did or if they were driving while using their phones.

— Kate Kaye (email| twitter)

A MESSAGE FROM UPWORK

Seeking to triple its employee base, Whisk, a fully remote team, sought diverse talent from a wide variety of regions through Upwork, a work marketplace that connects businesses with independent professionals and agencies around the globe.

Learn more

Nvidia’s data-center push gets some new hardware

Nvidia unveiled several new data-center chips and related tech Tuesday, including a new server CPU chip and its next generation of graphics-processing architecture for AI applications that will go on sale this year.

Nvidia’s Grace processor will feature two CPUs joined by a custom-designed interconnect intended to deliver high-speed data transfers with low latency. Based on the Arm v9 tech, the Grace CPU will have 144 high-performance cores — Nvidia claims it will double the performance and energy efficiency of existing “leading” server chips.

Nvidia unveiled its next-generation architecture called Hopper, and said it plans to sell the H100 server graphics processors based on it in the third quarter. The server chips will feature 80 billion transistors manufactured on TSMC’s four-nanometer fabrication process.

The Hopper-based chip includes a number of enhancements, including a natural-language processing engine, increased security features that allow for confidential computing and the ability to segment each GPU into smaller parts to handle different types of jobs.

— Max A. Cherney (email | twitter)

Deciphering ancient texts with AI

We know so much more about ancient history from actual texts that have survived thousands of years than you might think, but researchers believe we could unlock even more insights into the past given how many original texts exist but are unreadable thanks to the ravages of time. A group of researchers including scientists from Google’s DeepMind think they can use neural networks to fill in the missing gaps.

As detailed in Nature earlier this month, the researchers have developed a system called “Ithaca,” or “a deep neural network architecture trained to simultaneously perform the tasks of textual restoration, geographical attribution and chronological attribution,” according to the paper. Researchers using Ithaca to decipher texts in poor condition were able to improve their accuracy from 25% to 72%, which could produce an enormous gain in understanding.

— Tom Krazit (email | twitter)

Around the enterprise

Okta acknowledged that a hacker had taken control of an accountbelonging to one of its site responsibility engineers, which allowed them to gain access to internal Okta data but does not appear to have compromised customer data.

General Dynamics is not the first company many of you might think of when thinking about enterprise tech, but it just won a $4.5 billion deal to build a new hybrid cloud system for the National Geospatial Intelligence Agency.

HPE is one of those companies, and it launched a new version of its GreenLake hybrid cloud software focused on managing services.

Alphabet (aka Google, but fancier) spun out a new subsidiary called Sandbox AQ that will sell some marriage of AI and quantum computing services to “future proof their technologies.”

A MESSAGE FROM UPWORK

Whisk isn’t alone in unlocking the global marketplace to find the right types of employees to support its business goals. More than three-quarters of U.S. companies have used remote freelancers, according to research from Upwork, and more than a quarter of businesses plan to go fully remote in the next five years.

Learn more