Amanda Zantal-Wiener

one of the latest was last week's aws re:invent, where amazon announced a new suite of ai-enabled technology designed with businesses in mind.

and while these new products, features, and tools come with a host of opportunities for the developers, marketers, and others who plan to use them, they also raise a few questions. what are they designed to do? how can they help you? how do they work?

there's also one less-official question that occurred to me as i learned of these developments. is amazon trying to creep into google's territory?

let's take a look at some of these ai announcements from aws re:invent and dig deeper into just what they mean.

1. what is aws re:invent?

aws re:invent is what amazon describes as "a learning conference" produced by amazon web services -- that's what aws stands for. the intended audience is what it calls the "the global cloud computing community," but the event features content for anyone who wants to learn how cloud technology can help grow and scale a business. from adtech, to content delivery, to the internet of things, the conference ranges in learning opportunities from keynotes to certification sessions.

2. what were the major ai capabilities announced at aws re:invent 2017?

while a handful of new capabilities were unveiled by aws, there are four that i'd like to focus on (with a bit of teaser text below on what each one is capable of doing):

1. comprehend: text analysis for a variety of content formats
2. deeplens: a wireless, artificially intelligent video camera with newly enhanced object recognition
3. transcribe: speech recognition technology to convert the spoken word to text
4. translate: what it sounds like -- advanced text translation technology 

if some of these sound familiar -- or seem reminiscent of similar capabilities previously made somewhat famous by a certain search engine giant -- chances are, it's because they are familiar. i'll delve deeper into this below, but many of these capabilities reflect those for which google has been known to be a leader, especially in the realm of translation. after all, who could forget this jaunty promotional video on the topic?

3. what is deep learning vs. machine learning?

the reason why i want to establish the difference between these two technologies (and what they are, in the first place) is that most of these capabilities use one or both. 

machine learning essentially describes the ability of a machine to learn things -- habits, language, behaviors, and patterns, to name a few -- without having been programmed to do so, or pre-loaded with that knowledge. it doesn't describe artificial intelligence in entirety, but rather, is one very important ai capability.

deep learning is a type of machine learning -- and is a bit trickier to explain. basically, it takes the next (big) step in that it's designed to imitate the way the human brain works by way of something called neural networks. in our brains, we have biological neural networks in which the action of one neuron creates a series of subsequent actions that ultimately result in our different behaviors.

in technology, artificial neural networks seek to replicate that phenomenon by becoming "trained" to comprehend data based on a certain set of criteria. one of the more notable examples of deep learning in practice is in image recognition, in which neural networks are able to recognize different data patterns or cues to learn what, for example, a cat looks like.

4. what is amazon comprehend?

amazon comprehend uses something called natural language processing (nlp) to better understand and determine the meaning within text. it's an instance of machine learning, in which the technology learns how to comprehend -- if you will -- and process language as it was intended by the human being speaking or writing it.

comprehend performs this capability with a series of steps:

1. identify the language.
2. pick out the phrases that most strongly indicate what the text is about -- things like names of people, companies, places, or important events.
3. use those phrases to determine if the sentiment of the text is positive or negative.
4. file that set of text according to its topic within a collection of subjects it has already organized based on patterns it's observed.

so, how does this technology apply to the real world? well, it's particularly helpful in an instance of, say, analyzing written customer feedback. by feeding these comments to an api like comprehend, marketers can use this technology to synthesize data from their audiences to determine something like thematic areas of improvement.

5. what is deeplens?

simply put, deeplens is a high definition video camera that was designed with developers in mind. it was built with deep learning capabilities and what aws chief evangelist jeff barr describes as "pre-trained models for image detection and recognition."

in other words, it's a very smart camera: one that can recognize objects, faces, motions, and creatures (e.g., a dog from a cat). and while that's very cool -- not to mention, somewhat reminiscent of the recently-announced google clips camera -- there's a reason why it could prove so helpful to businesses.

to start, deeplens comes with a number of "templates," or recognition technologies that users can build upon for their own projects. object and action recognition, for example, can help to more seamlessly create something like product tutorials or demonstrations, by developing a system or algorithm that learns to recognize how the two are paired for different outcomes.

for example, if you're demonstrating how a certain cooking appliance can be applied to different scenarios, it seems that deeplens can be utilized in building a system to recognize the appliance itself (like a standing mixer), the actions the user can take with it (like mixing cake batter), and the resulting outcome (a delicious cake).

source: amazon web services

6. what is amazon transcribe?

anyone with a journalism background is more than familiar with the headache joy of transcribing spoken interviews. we want to get it just right, be sure not to misquote the interviewee, and communicate what was said in the right context.

if only, back in my earliest days of reporting, there were advanced transcription services available to the common writer.

but now, there's transcribe: an aws service that uses machine learning technologies to recognize the spoken word and transcribe it into text.

while the function itself is fairly intuitive, the benefits might not be. so let's lay out two instances where technology like this can be applied to a marketer's world:

1. interview transcription. i already covered this a few paragraphs earlier, but let's say you're writing up a blog post that requires quotes from a spoken interview. this technology eliminates the time-consuming step of transcribing what was said, and leaves you instead with all of the text from that conversation, allowing you to pick and choose the quotes you want to incorporate.
2. video transcription. accessibility is no longer optional. it's important to provide a way for individuals who are hearing impaired to be able to consume and enjoy your video content, even without the audio. transcribing what is said in the video by way of full paragraphs or subtitles allows them to do so -- and, it allows those who simply prefer to watch videos without sound to be able to absorb the content, as well.

these are only two of the more prominent examples of how such technology could be applied, but there are many more, from transcribing podcasts, to documenting notes from an important meeting.

7. what is amazon translate?

this development might be my favorite.

around here, we talk a great deal about approaching marketing with a global mindset. while i might use jetblue as a remarkable example of marketing, it might not resonate as much with audiences in countries where this airline doesn't operate.

to put it simply, the internet is a global, international destination. the people reading your content might not regularly engage with the same brands you do, and they might not speak the same language.

that's why a growing number of developers and marketers are building a multilingual web presence -- one where their online properties and content can be seamlessly viewed in the language preferred by the user. it's a trend that, as hubspot's own global presence continues to grow, i take inordinate glee in seeing.

it's also why i love seeing tools become available that make it easier to approach marketing with a "global first" mindset. translate is one such tool: a service that uses machine learning to more naturally translate text from one language to another. 

here's a look at how it worked when translating a french paragraph to english:

source: amazon web services

so, here's the million-dollar question: is amazon creeping into google's territory?

maybe.

it's not the definitive answer i hoped to have, but in these early release days, it might be too soon to tell. while most of google's headline-making ai developments are largely consumer-centric (like the previous example of clips), it is true that the company has been working on its own stack of machine learning capabilities for businesses. look no further than google.ai, for example, where the mission is to bring "the benefits of ai to everyone" -- including, i assume, marketers.

this is only the beginning.

what are you most excited about? what confuses or scares you, and what fills you with delight? feel free to weigh in with your thoughts on these aws ai developments on twitter, or let me know if you have a question about it.

originally published dec 5, 2017 8:00:00 am, updated december 11 2019

last tuesday, several amazon customers reported receiving an email from the online merchant reading, "someone great recently purchased a gift from your baby registry!" and while the internet typically can't be used for a pregnancy test -- unless you count target's 2012 public relations disaster after predicting a teen's pregnancy by tracking her shopping habits -- it still caused brief moments of panic among those who got the email.

there were some fears that the emails were a result of phishing attempts, but in the end, amazon confirmed to techcrunch that the emails were the result of a technical glitch, going on to send apology emails to the customers that received them. it's not clear what exactly happened or what the the glitch entailed, but let this be a lesson to marketers: triple check your email workflows.

among the panic, twitter had quite a bit of fun with the error:

more trouble for targeted ads

following last week's propublica revelation that facebook was allowing advertisers to use anti-semitic targeting criteria for promoted content, it was quickly discovered that google and twitter had similarly flawed advertising technology.

buzzfeed was the first to discover that the google allowed advertisers to use anti-semitic and racially-charged search terms to target certain audiences, and soon after, the daily beast reported that twitter allowed similar targeting criteria, which resulted in an audience of roughly 26.3 million users.

all three companies have since responded that they either have or are working to remove this criteria, with facebook coo sheryl sandberg publishing a very lengthy, formal apology on wednesday:

and more from facebook

yesterday, facebook released an official statement on its plans to hand over important documents regarding the ads it sold to russian organizations to congress, as well as a second one with answers to several "hard questions" on what happened to cause something like that to happen in the first place.

the gravity of that move is one that cannot be emphasized enough. as mike isaac writes for the new york times:

" ... the move to work with the congressional committees underscored how far the social network has strayed from being a mere technology company and how it has increasingly had to deal with the unintended consequences of the tools it provides to reach the more than two billion people who use the site regularly."

shortly after those posts went live, mark zuckerberg delivered a live address on his own page to address the efforts it would make moving forward to "protect the integrity of the democratic process."

some believe that this address, along with sandberg's statement form the previous day, is the first of many efforts by facebook to proactively dodge federal regulation by staying one step ahead of congressional actions or attempts to curb what such channels and platforms can actually do. it's even, perhaps, a defensive move, as the legality of the aforementioned ad sale remains in question.

meanwhile, twitter is also due to appear at a senate intelligence committee briefing nxt week to further examine its own possible role in influencing the most recent u.s. presidential election.

the uber-alphabet lawsuit got even messier

last week, we filled you in on the ongoing lawsuit between uber and alphabet, inc company waymo over proprietary self-driving technology. since then, there have been a few key developments.

first, over the weekend, alphabet requested that the trial be postponed after receiving crucial information that the court ordered uber to turn over. seeing that information, it seemed, made waymo realize just how much was at stake with uber being in possession of these materials, and needed more time to review all of the evidence supporting its case. megan rose dickey of techcrunch tweeted a key portion of its statement on the issue:

waymo wants to postpone the trial w/ uber. here's waymo's statement. pic.twitter.com/sf8crgdyak

— megan rose dickey (@meganrosedickey) september 17, 2017

uber, of course, objected to that request, charging that alphabet is trying the delay the trial -- december 5 is the proposed postponed date -- not because of surmounting evidence, but because of a lack of it. the full opposition can be found here:

that same judge also gave uber permission to publicly disclose some of what waymo is hoping to gain from the lawsuit: $2.6 billion for one stolen (allegedly) trade secret. but there are still eight other secrets that waymo says to have been stolen by uber, and no monetary figure has yet been assigned to them.

at this point, the trial is still set to begin on october 10, and recently-appointed uber has a decision to make: whether or not he wants to settle out of court, or continue to defend the company's name in what promises to be a complex, drawn-out trial.

it's just one of many problems for uber these days. with the release of ios 11 this week, uber was forced to allow users to block the app from tracking their locations.

additionally, the bbc broke news this morning that transport for london would not renew the ride-sharing app's private hire license, calling it "not fit and proper" to carry on operations there. uber has 21 days to dispute that decision and can continue providing services in london until then.

attack of the flying eggplants

i'll admit it -- my new favorite feature of ios 11, the latest operating system available on the iphone, is probably the ability to fill your imessage recipient's screen with the next, image, or emoji of your choice. but just for the sake of due diligence, i tested it by sending this gem to one of my colleagues:

but my low bar for amusement aside, the new operating system comes with some features that are actually, you know, productive. here are our five favorites:

1. screen recording. so, just how did i capture the magical moment above? ios 11 has a screen record tool that saves the video in your camera roll.
2. do not disturb while driving. this "do not disturb" feature uses your car's bluetooth connection to turn on automatically while you're driving.
3. screenshot tools. the new ios has a brilliantly simple new screenshot feature, which lets users draw on, crop, or highlight with ease.
4. gifs in camera roll. the camera roll now allows users to save and view gifs, plus the newest editing tools even enable you to turn your live photos into gifs.
5. notes app upgrade. the notes app now features useful tools like a document scanner, and the ability to insert all kinds of formatting into your note.

and, finally -- we can't forget arkit -- apple's mobile augmented reality technology -- which has been a big portion of the talk of the iphone town in the days following ios 11's release. i tweeted about my experience with using it on wayfair's home shopping app:

so, i don’t think this is gonna fit. #arkit pic.twitter.com/arpsnr1cib

— amanda zantal-wiener (@amanda_zw) september 20, 2017

what's google up to?

a lot happened for the search giant this week, beyond its parent company's lawsuit and a significant team acquisition. first, there were some leaks around the rumored october 4th release of the pixel 2 and pixel xl, but they were mostly limited to the device's available colors, as per droid life.

that same outlet also leaked the rumored google home mini, a much smaller version of the google home, which many are calling the company's response to the echo dot. and on tuesday, a “media streaming device” -- the same language used to describe the original google home -- with features remarkably similar to the first google home was submitted to the fcc. these developments all align with the timeline leading up to the october 4th press event.

source: droid life

in non-pixel or home news, google announced four new features this week:

1. the google app on ios will now have a suggested content feature, in which users are provided with visual links to more information on what they're reading about.
   source: google
2. contact information like phone numbers, addresses, and email addresses will now -- finally -- be automatically converted into hyperlinks on gmail. official announcement here.
   
   source: google
3. natural language processing has been added to cloud search -- a search tool within g suite -- to help users more quickly find information based on the words that google deemed to be the most frequently used among g suite customers: "what," "who," "how" or "when." for example, if you remember which one of your colleagues sent you a shared document, but can't remember what it was called, now you can search for it with a query like, "docs shared by karla."
   
   source: google
4. read receipts are coming to gmail -- kind of. on wednesday, google announced the launch of email log search, which allows g suite users to track the status of sent emails, such as where it is (e.g., the trash), or if it's been opened.
   
   source: google

actually, no, equifax still isn't handling this data breach well

remember all of those marketing takeaways from the equifax hack that we outlined last week? well, it turns out that equifax hasn't exactly heeded that advice -- or that of too many others, it seems. in fact, it was revealed earlier this week that the company's customer service agents on twitter were directing customers to a fake website that, visually, was nearly identical to the site equifax set up for users to enroll in free credit marketing.

the clone site was created by by software engineer nick sweeting, whose intentions weren't malicious, but rather, were to show how poorly equifax was monitoring and managing the situation.

not only did they tweet the wrong link, they tweeted it 3 times. #equihax pic.twitter.com/t8jrhsfhqw

— nick sweeting 🚲 (@thesquashsh)   september 20, 2017

sweeting was quite transparent about that in creating the site, which has since been taken down, along with any tweets directing customers to it -- it was titled, "cybersecurity incident & important consumer information which is totally fake, why did equifax use a domain that’s so easily impersonated by phishing sites?”

the "fake" site did not collect any personal information, but sweeting pointed out how easily it would be for other hackers to create an equally identical site that did using the linux command "wget" -- and he blamed that on equifax's choice to establish an entirely new domain, rather than create an equifax.com subdomain.

"wget" essentially permits anyone -- yes, anyone at all -- "to just suck their whole site down with wget and throw it on a ... server,” sweeting explained in an email to the new york times. his version, he said, had "the same type of ssl certificate as the real version, so from a trust perspective, there’s no way for users to authenticate the real one vs. my server.”

creating a subdomain should have been the obvious move for equifax, carnegie mellon is professor rahul telang told the outlet, "so that if somebody tries to fake it, it becomes immediately obvious.”

this development comes amid news that equifax actually suffered more than one hack this year. in addition to the headline-making breach in july, the company experienced an earlier one in march, creating even more confusion around the decision to wait until september to alert customers, as well as the massive august stock sale by its executives.

odds and ends

i don't want to grow up, because that means i have to file for chapter 11

for anyone who grew up begging their parents to take them to the toy store, this week came with some sad news: toy store chain toys ‘r’ us filed for chapter 11 bankruptcy protection this week. many were quick to blame its demise on amazon, which has been named the culprit for the financial woes of many other brick-and-mortar retailers, but in reality, the cause may have reached far beyond that. as recode reports, the move is largely the result of a "cocktail" of limited product selection, a lack of competitive pricing, and piling debt after several 2005 buyouts.

for the sake of our own childhood memories, we hope toys 'r' us is able to turn things around.

the dhs got served ... by an association of vcs

allow us to introduce you to the international entrepreneur rule: a federal measure that, had it passed in july 2017 as planned, would have made it easier for foreign entrepreneurs to obtain visas for the purpose of founding startups in the u.s.

however, the same month it was slated to be effective, the current presidential administration delayed it until march 2018, with many believing that it will only go on to be completely dismantled. but this week, axios reports, the national venture capital association (nvca) has brought forth a lawsuit against the department of homeland security, on the grounds that the decision to delay the enforcement of the rule violates the administrative procedure act -- which says that the department must first "solicit public comment."

the formal complaint can be downloaded here.

no empty nest here

nest, which was acquired by google back in 2013, held its first major press event this week, where it unveiled a number of new products. among them were the nest cam iq -- this writer's personal favorite unveiling from the launch -- an outdoor security camera that can detect movement and differentiate whether it's coming from a person or an object. if it's a person, the system alerts you, as well as letting you know if it senses a barking dog or a talking person. even better: it's equipped with facial recognition, so that if someone familiar comes into the camera's range, like your regular dog walker, the system will recognize that it's likely not an intruder.

also announced was the nest hello video doorbell, which uses similar camera technology to the above to alert users if a person is within range, even if they don't ring. finally, an overall comprehensive security system was unveiled called nest secure, which exists of three key components: nest guard, where the system is armed and disarmed with the second component, nest tag, which is similar to a key fob and can be used to turn off the alarm system. the first piece is nest detect, which can sense general motion and the opening or closing of windows or doors.

check out the video summary here:

that's all for this week! next week, we're off to inbound 2017: one of the world's largest and most remarkable marketing and sales industry events. we'll be back with our regular news coverage the first week in october.

until then -- happy autumn.

originally published sep 22, 2017 7:00:00 am, updated september 22 2017

google is no stranger to algorithmic change. and usually, those changes are made for the sake of the user. looking at a history of google's product announcements, usability is usually at the heart of the modification.

so when google announced its impending smackdown on mobile pop-up ads earlier this week, it came as no surprise that the major reason behind it was to enhance the user
experience.

for many businesses, the announcement carries major implications. those that rely on advertisements as a primary source of revenue, for example, will be some of the hardest hit. to help you navigate this change, we put together everything you need to know below, from what the update entails to how to prepare accordingly.

what’s new in mobile search results

back in 2014, google added a "mobile-friendly" label for search results that were optimized for such platforms -- where text is readable without zooming or horizontal scrolling, and links are spaced well enough so that there’s a reduced chance of mis-tapping.

but two years later, google has found that 85% of mobile search results are optimized that way. as a result, the search engine is doing away with that label, and introducing new mobile-specific ranking criteria.

and now, pages with mobile pop-ups -- or what google is calling “interstitials” -- probably won’t be ranking as highly when these algorithmic changes take effect on january 10, 2017.

“pages that show intrusive interstitials provide a poorer experience to users than other pages where content is immediately accessible,” google’s official announcement states. “this can be problematic on mobile devices where screens are often smaller.”

not all pop-ups are created equal, though, so there are some specifics around which types of interstitials google considers to be disruptive to the user experience. some are legally required -- like ones used by liquor companies that verify the user's age -- so they won't impact the page's rank.

according to the official statement, interstitials affected by google's crackdown include the following:

• "showing a popup that covers the main content, either immediately after the user navigates to a page from the search results, or while they are looking through the page.
• displaying a standalone interstitial that the user has to dismiss before accessing the main content.
• using a layout where the above-the-fold portion of the page appears similar to a standalone interstitial, but the original content has been inlined underneath the fold."

what does it mean for marketers?

as we mentioned before, the companies that rely on these interstitials for income will be especially impacted by this change. they're the ones who, as emma hinchliffe of mashable points out, need that ad revenue to survive. now, these businesses face a difficult choice: rank, or profit.

but losing seo traffic can "crush" these companies, says hubspot's senior product marketing manager, marcus andrews. and that makes sense -- 51% percent of digital media is consumed via mobile, compared to a lagging 42% on desktop.

andrews suggests that, if they haven't done so already, marketers solve for mobile seo first. the pain that comes with changing a revenue model is inevitable, but shorter-term -- and businesses that rely on advertiser dollars, he says, should figure out ways "to make money that don’t totally disrupt the mobile user experience."

"google is very focused on the user," andrews continues. "marketers are always looking for hacky ways to increase traffic and conversion rates, and every once in a while, google needs to make a correction to improve the user experience."

that actually creates a great opportunity for marketers to think more about the user -- both the experience, and what that person is offered. it's what hubspot's director of product development, nicholas holland, calls a "forcing function." it makes marketers seriously consider the increasing overtaking of mobile technology, and what the implications will be on their overall business operations.

basically, these developments from google are a giant wake-up call to those who "create a bad viewer experience," holland says -- especially those who might not even realize it. now, they absolute must "think through alternative revenue methods."

but what are those methods, exactly?

first -- if it isn't obvious by now -- remove any pop-ups you've been using that are interruptive in nature. (read this post to learn what makes a pop-up not interruptive.)

"as inbound marketers, we rely on driving relevant visitors to content," says hubspot's principal product marketing manager, jeffrey vocell. "interstitials, especially interruptive ones, do not provide a good experience, and in many cases actually block or limit the content that can be seen."

that's why your best option might be to create valuable content that draws the visitors that vocell is alluding to. when you do that, holland advises, you can focus on driving revenue -- or at least leads -- using calls-to-action and embedded forms. (and to learn more about converting those leads, check out our free ebook on optimizing landing pages.)

replacing intrusive interstitials with valuable content is a double-edged seo sword. not only are you giving the user what he or she is searching for -- and improving your rank accordingly -- but you're also getting rid of the invasive pop-ups that, come january, would be lowering it.

what you can do now to prepare

if you're freaking out about google's announcement, that's okay -- but please, don't be. as we mentioned, these changes actually provide a great opportunity to use inbound marketing to generally enhance your marketing presence -- on mobile, or otherwise.

here's what you can do to get ready for the rollout:

• ditch your interruptive interstitials -- unless they're required by law. those include age verification displays, as noted above, and pop-ups that let your user know you use cookies.
• if you relied on interstitials for ad revenue, figure out where that money is going to come from now.
• find ways to generate revenue without obstructing the user experience, and in a way that optimizes your page for mobile. both of those factors will likely remain crucial to search engine rank.
• know that those solutions often exist in the content you create. make something valuable for the user. by gating it behind a landing page, you're generating leads -- and eventually sales -- in a much less intrusive way that brings visitors to you.

what do you think about google's latest announcement, and what are you doing to prepare for it? let us know in the comments.

 

originally published aug 25, 2016 8:00:00 am, updated december 11 2019