Tech Brew Ride Home - Thu. 04/11 - Alexa Snags Its Own "Scandal"

Episode Date: April 11, 2019

Yes, there are people listening in on your Alexa recordings, but it’s complicated; Jeff Bezos challenges competitors to pay their workers more, the third big unicorn IPO of the year happened today, and a rundown of noteworthy recent startup raises. Sponsors: Tiny.website LogiAnalytics.com/ride Links: Amazon Workers Are Listening to What You Tell Alexa (Bloomberg) Jeff Bezos challenges retail rivals to outdo Amazon’s $15 minimum wage (The Verge) YouTube TV Hikes Price to $50 per Month for All Customers After Inking Discovery Deal (Variety) PagerDuty pops more than 50% in debut as tech IPO market heats up (CNBC) Affectiva raises $26 million to bring emotional intelligence AI to car safety systems (VentureBeat) Triplebyte raises $35M for its online coding test and credentialing service for hiring engineers (TechCrunch) Lemonade picks up $300 million Series D led by SoftBank Group (TechCrunch) That image of a black hole you saw everywhere? Thank this grad student for making it possible (CNN) Subscribe to the ad-free feed RIGHT HERE inside your podcast app Learn more about your ad choices. Visit megaphone.fm/adchoices

On April 4th, 2023, around 2 in the morning, a man was found stabbed multiple times on a sidewalk in downtown San Francisco. Hey, who did this to you? What happened next turned the story into a political firestorm. Reports have identified the victim as Bob Lee, the founder of Cash App. From Bloomberg Podcasts, this is Foundering, the Killing of Bob Lee, beginning April 16. Welcome to the TechMeme Ride Home for Thursday, April 11th, 2019. I'm Brian McCullough today. Yes, there are people listening in on some of your Alexa recordings, but it's complicated. Jeff Bezos challenges competitors to pay their workers more. The third big unicorn IPO of the year happened today and a rundown of noteworthy recent startup raises. Here's what you missed today in the world of tech. Bloomberg has a piece up that describes Alexa's AI training process, whereby actual human workers transcribe voice clips generated by echo devices. Those workers can apparently see customer account numbers related to the echo devices and the audio they're

transcribing and hear private conversations. Quoting Bloomberg, Amazon employs thousands of people around the world to help improve the Alexa digital assistant powering its line of echo speakers. The team listens to voice recordings captured in Echo Owners' homes and offices. The recordings are transcribed, annotated, and then fed back into the software as a part of an effort to eliminate gaps in Alexa's understanding of human speech and help it better respond to commands. The Alexa voice review process described by seven people who have worked on the program highlights the often overlooked human role in training software algorithms. In marketing materials, Amazon says Alexa, quote, lives in the cloud and is always getting smart. But like many software tools built to learn from experience, humans are doing some of the teaching, end quote. So Alexa allows Amazon contract employees to listen in on what you're saying in your home.

That's kind of the long and short of it, and that's sort of a scary elevator pitch headline generated the expected level of outrage on Twitter and elsewhere online. However, and I'm not trying to be a contrarian here, should anyone be surprised by this? As Ben Thompson pointed out in his Stretechory note this morning, nine paragraphs into the story on Bloomberg, there is this. Quote, we have strict technical and operational safeguards and have a zero tolerance policy for the abuse of our system. Employees do not have direct access to information that can identify the person or account as part of this workflow. All information is treated with high confidentiality, and we use multi-factor authentication to restrict access, service encryption, and audits of our control environment to protect it, end quote. That's from an Amazon spokesperson. Also, 19 paragraphs down in the Bloomberg piece, there's this, quote, Apple's Siri also has human helpers who work to gauge whether the digital assistance interpretation of requests lines up with what the person said.

The recordings they review lack personally identifiable information and are stored for six months tied to a random identifier, according to an Apple's security white paper. After that, the data is stripped of its random identification information, but may be stored for longer periods to improve series voice recognition, end quote. So what I'm saying is this. Maybe people have to have some sort of a deeper understanding of how AI stuff works going forward. The robots are learning on their own to an increasing degree, but at this point, most of AI stuff is still human beings training the computers and doing so very much by hand. Why do you think all the self-driving car companies spend all this time driving all those miles out on real-world roads? They are hand-training the algorithms and the systems, etc. As Ben Thompson says in today's strategicry note, quote,

How else can a speech-to-text algorithm, which is at the core of Alexis service, improve if not by leveraging entities that have organic, highly developed speech-to-text capabilities, that is to say, humans, end quote. As Ben says, heck, even Apple, privacy-focused Apple, is doing something similar, though it should be noted that they're doing it with a higher degree of, privacy safeguards built in. Now, on the other hand, as Owen Williams pointed out in his note this morning, quote, Amazon says that only a small amount of audio data goes through this process, but it doesn't really allude to humans reviewing it anywhere in its FAQ or privacy sections. It has always implied that machines are behind the magic, but the reality appears pretty different. There's a bunch of manual processing going on, end quote. And this is from Owen again, and I think this part is key, quote, voice data is a data treasure trove and a competitive advantage in the future.

Amazon, Google, and Apple sit on vast supplies of audio files from users speaking to their gadgets. But what happens inside that black box remains a mystery for the most part, and perhaps a liability in the long term if humans are accessing the data, end quote. So to me, if you want to get outraged to the degree that this outrages you, the scandal here, if there's is one, is somewhat akin to the scandal again, if you think it is a scandal, that all of the image recognition AI, for example, has been trained using pictures of our faces, and none of us agreed to give up our faces as grist for the mill. Speech recognition AI needs a humongous corpus of audio clips to get better. Is it fair that Amazon is in a position to just take it and hoard it and most importantly take it from us without compensating us for it, extract value from

us without asking permission? Is it fair that the big oligops are the ones who will be sitting on this pile of hugely valuable data, which, as Owen pointed out, will be a big competitive advantage for them going forward? I don't know. That's where the real question lies to me. Maybe AI really is a zero-sum game, and there is a land grab going on, and the leaders today might be the leaders forever because they got there first, and that might suck. But come on, in this day and age, you put a microphone in your home. Yes, someone is going to be listening to you on that microphone somewhere. If you're still in denial about the base reality of that, then there's a bridge right outside my window that I'd be happy to sell you. In his annual letter to Amazon shareholders,

Jeff Bezos challenged Amazon's retail rivals to match or beat Amazon's recent move to increase its minimum wage to $15 an hour. Quote, today I challenge our top retail competitors, you know who you are, to match our employee benefits and our $15 minimum wage, Bezos wrote. Do it. Better yet, go to $16 and throw the gauntlet back at us. It's a kind of competition that will benefit everyone, end quote. Now, we've spoken about this before. This is not just altruism on Amazon's part. It's actually smart strategy. It would only benefit Amazon if labor costs at Walmart and Target, etc. were forced upwards. Your shrinking margins are his opportunity or something like that. Still, I have to admit, it would be nice to see some sort of a wage war break out in an upward

direction in commerce that would actually benefit workers. Target apparently plans to bump its minimum wage to $13 an hour up from the current $12 an hour and hopes to move to $15 an hour by the end of 2020. Walmart's minimum wage is still $11 an hour. But Costco, by the way, announced a move to $15 an hour just last month. YouTube TV is raising its price to $50 a month across the board for all customers, which means a 25 to 43% increase in price depending on what package you signed up for. Quoting from Variety, the new pricing kicks in after YouTube signed a multi-year carriage deal with Discovery for YouTube TV, covering linear programming and access to nearly 50,000 on-demand titles.

On April 10th, YouTube TV will launch eight Discovery-owned networks, Discovery Channel, HGTV, Food Network, TLC, ID, Animal Planet, Travel Channel, and Motor Trend. Own, the Oprah Winfrey Network, will launch on YouTube TV by the end of the year, end quote. A couple of observations about this. First, we thought that Netflix rivals might come after Netflix by undercutting them on price, and so far not so much. But second, when we talk about blowing up the TV cable bundle just to reconstruct it all over again, Well, here you have exactly that.

You have a streaming over-the-top video service that is jacking up prices, likely just to cover the carriage fees that it had to pay in order to offer a slew of channels from a content network. That's exactly the way the legacy cable bundle operated. The third tech unicorn IPO of the year happened today, and so far so good, at least at the time of this writing, quoting from CNBC, pager duty shares jumped more than six. 60% in their opening day of trading on Thursday in the first notable software IPO of 2019. Shares sold around $39 a piece, leaving PagerDuty with a market capitalization of around $2.7 billion. The stock had been priced at $24 a share Wednesday at a market cap of nearly $1.8 billion, end quote.

I probably don't have to tell you nerds this, but in case you're not familiar with PagerDuty, it's a dev-ops software maker, which helps dev-es and technical teams, quickly spot problems and applications and respond to all sorts of user incidents. Pager duty is used by 11,000 companies including Slack, Box, and Netflix, and probably by a lot of you listening on a daily basis. Revenue was up 48% last year to $117.8 million, though it did record a $40.7 million loss as well. The dev-ops market is expected to reach $10.3 billion a year by 2023, up from $3.4 billion last year. PagerDuty was founded in 2009 by former Amazon developers. The biggest investors are Andreessen Horowitz, which owns 18% of the company, followed by Excel at 12%. And PagerDuty was a graduate of Y Combinator's summer 2010 class.

Which reminds me, I haven't done that thing where I cover a grab bag of interesting startups who've made headlines by raising money recently, so let's do that. Effectiva uses AI to detect emotion in vehicle passengers, quoting from Venture Beat. Effectiva's computer vision system may recognize a driver is drowsy, then use text-to-speech AI to ask the driver if they would like some music to change the temperature or to recommend the car be pulled over, end quote. Other customers for Activa's tech include ride-sharing companies that could use the computer vision to recognize, say, if you leave your bag in a car or detect a poor customer experience in your latest ride hail based on reading sentiment analysis of your face when you're a writer.

Effectiva just raised $26 million led by automotive supplier active. Schedule spelled S-H-E-D-U-L is an online booking platform for salons and spas. It just raised a $20 million series B. be at a $105 million evaluation. Armis helps companies protect Internet of Things devices on their networks, and it raised a $65 million Series C led by Sequoia Capital. Triple Buy offers personalized online coding tests and technical interviews to help employers screen candidates, quoting from TechCrunch.

LinkedIn, with nearly 600 million users, dominates the market today for recruiters headhunting suitable candidates for knowledge worker roles. But with minimal controls for guaranteeing that people are what they say they are, it opens up a gap in the market for startups to give recruiters a better steer with more verified information. Triplebyte, which offers personalized online coding tests, the test asks you questions based on how well you have answered previous problems, and subsequent technical interviews to help screen candidates for prospective employers, is announcing that it has raised a Series B of $35 million underscoring the demand it has seen for its services and the opportunity to grow the business further, end quote. And finally, Lemonade went the Mega Raise route, raising a $300 million series D led by who else SoftBank.

Lemonade is an insurance startup that uses AI-powered bots to digitize the insurance buying experience for home renters and homeowners. tech crunch once again. Users simply download the app and answer a few questions before getting a quote which starts at $5 a month but can surely go up based on a number of factors, including how much personal property one owns. The company has also differentiated itself from traditional insurance providers by integrating a give-back system directly into the product. Lemonade takes a fixed slice of users' monthly payment as revenue and sets the rest aside for claims. Unclaimed premiums, go to the user's charity of choice. company has grown significantly since launch, last year hitting 57 million in revenue.

Co-founder and CEO Schreiber says the company is on track to do 100 million in revenue this year and that they've sold 500,000 policies to date, end quote. Of course, insurance is a capital-intensive business to get into, thus the need to raise to the tune of hundreds of millions of dollars. More on the concept of mega-rounds on one of this weekend's bonus episodes. Finally today, you've probably seen her picture making the rounds yesterday, as it's sort of gone viral. But one of the scientific heroes from yesterday's announcement of the first picture of a black hole is MIT grad student, Katie Bauman, who helped develop the imaging algorithm for the Event Horizon Telescope that captured the now iconic picture of the black hole, which everyone has now memed into crispy cream donuts or the eye of Sauron or whatever. Using imaging algorithms developed by Bauman and others, researchers created scripted.

code pipelines to stitch the photo together from tons of different sources, quoting CNN, quote, they took the sparse and noisy data that the telescopes spit out and tried to make an image. For the past few years, Bauman directed the verification of images and selection of imaging parameters. We developed ways to generate synthetic data and used different algorithms and tested blindly to see if we can recover an image, she told CNN. We didn't want to just develop one algorithm. We wanted to develop many different algorithms that all have different assumptions, built into them. If all of them recover the same general structure, then that builds your confidence. The result? A groundbreaking image of a lopsided ring-like structure that Albert Einstein predicted

more than a century ago in his theory of general relativity. In fact, the researchers had generated several photos, and they all looked the same. The image of the black hole presented on Wednesday was not from any one method, but all the images from different algorithms that were blurred together, end quote. Again, there were a ton of scientists on this project, but the image of the image imaging portion was led by junior researchers like Bauman, 29, who starts teaching as an assistant professor at California Institute of Technology in the fall. As Alex Tausig joked on Twitter, Hello, fellow grad student. What's your thesis on? Katie Bauman. Oh, I'm writing an algorithm that will construct the first ever image of the black hole at the center of our galaxy,

directly confirming 100 years of physics theory. Oh, Congratulations, Katie and the whole team. That's all for today. Some more questions have trickled in for the listener Colin episode. So keep those coming if you're interested, and maybe we'll be able to put that together for next weekend. I just locked and loaded the second interview for this weekend's bonus episodes, though. So two excellent interviews are coming your way in a couple of days. As always, I've been your host, Brian McCullough, at Brian MCC,

on Twitter. The subreddit for the podcast is R-slash Ride Home, where you can tip me stories. And if you felt like supporting the show directly, the link to the ad-free premium feed is at the bottom of the show notes. Talk to you tomorrow.