Good morning. This week = earnings bonanza for the tech industry. We’re deploying a virtual assistant to track how many times execs say “[X] trend has accelerated” or “digital transformation.” The Brew’s current over/under is set at 88.5 mentions. —RD
In today’s edition:
Smart assistants Bipedal bots RIP
—Hayden Field, Ryan Duffy
|
|
Francis Scialabba
It’s hard enough for humans to keep up with what the cool kids are saying—so imagine how much tougher it is for AI.
Emerging Tech Brew chatted with Nino Tasca, Google Assistant’s group product manager for speech, about the team’s strategy for addressing an obstacle common across the natural language processing (NLP) world: understanding the ever-changing nature of language.
Lost for words
Slang, cultural references, accents, different dialects...attempting to juggle all of these things is like playing whack-a-mole. Machine learning models have to simultaneously play catch-up and stay ahead of the curve.
- One secret: training the model on top music charts, which belt out new slang at full volume.
As for learning an entirely new language? Before any launch, Tasca says his team has to decide whether the model grasps five overall principles: linguistic accuracy; cultural relevance; how to sound natural; cultural norms; and voice, tone, and slang.
Case study: When “Despacito” blew up in 2017, Assistant’s model was less prepared to handle two languages at once than Justin Bieber was when he performed the song. But Tasca says the increased user interest helped accelerate the model’s learning how to handle multilingual user queries.
Custom-made
When learning a new language, customs and cultural references are just as important as words, tense, and sentence structure.
That’s why Tasca’s team tests any new launch in a company-wide internal program first—and why Google hired people from the 90+ countries where Assistant is available, who can help catch bugs or issues that non-native language speakers could miss.
Case study: Google Assistant largely understood the names of American sports stars, but a European team member realized it couldn’t recognize an international table tennis champion.
- “It was simply a question most of us in the U.S. had never tried up until that point,” says Tasca. “This realization sparked more effort into understanding uncommon words.”
Zoom out: The obstacles faced by Google Assistant are common to virtually every consumer-facing NLP product (think: Amazon’s Alexa, Apple’s Siri, Microsoft’s Cortana, and more). But there’s a straightforward solution: hiring a diverse team to build and maintain your NLP product/service.
|
|
Agility
On Thursday, Agility Robotics announced a $20 million in new funding. The Oregon startup plans to scale manufacturing of its bipedal robots for logistics, retail, and e-commerce clients.
“We’re having a bit of a return to U.S.-based manufacturing,” CEO Damion Shelton told the Brew. “The older arguments about cost have taken a backseat to quick iteration and more advanced manufacturing techniques.”
Last year, Agility launched Digit, a robot with two legs, two arms, an upper torso, sensors, and a tiny lidar head. Unlike many of its commercially available brethren, the robot is built to operate in human environments...like the front porch.
- Ford said in January that it would be Digit’s first customer. The company plans to eventually use Digit for autonomous first- and last-mile package delivery.
- Shelton said Agility is exploring non-logistics use cases like telepresence and “virtual tourism.”
Next steps: Agility is following a common line in the bot biz. It will initially focus on industrial “back-end work” and jobs that are “dull, dirty, and dangerous,” Shelton said.
|
|
Here at the Brew, our sales team has moved from in-person meet-and-greets to 50 Zoom calls a day. And it isn’t always easy to build a relationship without personal interaction (or a free lunch).
So we put together an article with the digital masterminds at DocuSign to discuss how our team uses DocuSign’s tech to sell efficiently when things continue to change every. single. day.
Featuring our very own Brand Partner (and fun fact: DJ), Jenna Castaldo, we’re getting the low-down on how she creates meaningful relationships that drive revenue.
So if our brave new digital world has made you yearn for the sales calls of old, take a read of our article with DocuSign. It’ll teach you the ways of automated contracts, eSignatures, and countersigns with incredible speed.
Check out our article with DocuSign right here.
|
|
Jung Yeon-Je/Getty
Lee Kun-hee, Samsung Group chairman for three-plus decades, died yesterday at 78. He’s credited with transforming the chaebol—a large, family-run firm—into a global tech powerhouse.
Samsung is now a top manufacturer of semiconductors, smartphones, and consumer electronics. But the conglomerate does much more, with subsidiaries in industries ranging from shipbuilding to insurance sales. Fun fact: Samsung built the world’s tallest building, the Burj Khalifa in Dubai.
Some South Koreans call their country “The Republic of Samsung.” Since Samsung contributes roughly 17% of South Korea’s GDP, the moniker isn’t misleading.
- Lee told employees he was driven by a constant sense of crisis—to keep the company innovating and ward off complacency.
- His tenure at Samsung Group was not controversy-free. In 1996, Lee was convicted of bribing politicians, and in 2008, of tax evasion. He received presidential pardons for both incidents.
Bottom line: Even though he was out of the limelight in recent years, Lee played a pivotal part in Samsung’s continued evolution. Investors are watching how Samsung and Lee’s heirs will pay gargantuan inheritance taxes on his shares.
|
|
Nissan
Stat: Japan, the world’s third largest economy and fifth largest C02 producer, will target zero emissions by 2050. EV manufacturers = .
Quote: “I think it’s possible to look at technology in the same way that Marxists would talk about history, as this thing that has its own agency and force and inevitability. It’s this thing that’s just happening, and you’re moving with it.” —Moxie Marlinspike, founder of Signal, to the New Yorker.
Read: The Markup compiled a list of tech issues on the ballot, including the gig economy, warrantless phone searches, privacy, facial recognition, and publicly funded broadband.
|
|
SPONSORED BY AUTONOMOUS SMART DESK
|
At least our desk is smart. While we might be prone to the occasional mistake, using the Autonomous SmartDesk 4 makes us feel like geniuses. It’s the first ever anti-collision user-adjustable desk that integrates with the Autonomous app for even greater control. And in case the strong, silent type is your type: SmartDesk 4 holds up to 350 pounds and it’s quieter than a cute lil’ bird call. Learn more about the SmartDesk 4 here.
|
|
-
Apple is developing third-gen AirPods and a second iteration of AirPods Pro, sources tell Bloomberg. Other Apple audio products, like over-ear headphones and a refreshed HomePod mini, are also coming down the pike.
-
ByteDance is mulling a Hong Kong listing for Douyin, the Chinese version of TikTok, Reuters reports.
-
Oculus owners will lose store purchases and account info if they delete their Facebook profiles.
-
Fitbit has 500,000 paid subscribers, CEO James Park tells the WSJ.
-
Weibo users say censorship is accelerating on the Chinese microblogging site.
-
Internet pioneer Vint Cerf is working on an interplanetary networking protocol.
-
Facebook is calling Irish content moderators back to the office after they were classified as essential workers, per The Guardian.
|
|
Monday: NASA announces an "exciting new discovery" about the moon; Earnings (Twilio, Hyundai).
Tuesday: Earnings (Microsoft, AMD).
Wednesday: Jack Dorsey, Mark Zuckerberg, and Sundar Pichai testify at a U.S. Senate committee hearing on Section 230; Earnings (Pinterest, Sony); Lenovo Tech World runs through Thursday.
Thursday: Earnings….deep breath...for Samsung, Facebook, Apple, Amazon, Google, Twitter, Spotify, and Shopify.
Friday: LG earnings; Halloween Eve.
|
|
For AI artwork: Web developer Matt Round created a generative adversarial network (GAN) and trained it on hundreds of street art photos of “a certain” street artist’s works. The AI, named GANksy, released a collection of 256 original pieces. View the gallery here.
For the Monday scaries: Killed by Google cheekily tracks discontinued and deprecated products from you can probably guess who. The site’s Twitter account is currently stylized as “Killed by Ghoulge .”
For a McFlurry status report: McBroken tracks whether McDonald’s U.S. ice cream machines are working. The site’s creator, 24-year-old Rashiq Zahid, says he reverse-engineered McD’s internal API and sends an order worth ~$19,000 every minute to check on the machines.
|
|
Catch up on the top Emerging Tech Brew stories from the past few editions:
|
|
Written by
@haydenfield and @ryanfduffy
Was this email forwarded to you? Sign up here.
|
ADVERTISE // CAREERS // SHOP
Update your email preferences or unsubscribe here.
View our privacy policy here.
Copyright ©2020 Morning Brew. All rights reserved.
40 Exchange Pl., Suite #300, New York, NY 10005
|
|