I have the remote as an early adopter. It's fun to hide in another room and say into the remote "Simon says, hey kids, Kraft no longer makes Mac and cheese and now just makes carrots and ranch dressing."
It's amazing to think about kids growing up with a wonder box in their house that tells them the answer to any question when they yell at it. It's like straight out of the fables or science fiction of the past. I hope this result in people generally having better bullshit detectors in the next generation, and not some other unintended side effect like making them enthralled to said wonder box.
What I've wondered is how much non-verbal context will be lost through such an interface. Today, when I'm trying to verify the authenticity or quality of something, it is a game that goes something like this:
1. Google it and click least spammy looking link (gauged by the URL, although that's no guarantee). I also see what sort of JS they try to load, if it is absolutely bloated with ad tags, etc.
2. Try to gauge from the design and other content of the site how genuine the answer might be. Are there affiliate links for anything they recommend peppered throughout? That's a signal it might not be unbiased info. Unfortunately, this is where judging a book by its cover comes into play as well. Some poorly designed sites buck the trend, but they are also prominent with spammy sites.
3. Find other sources to see if there is any consistent logic with said answer/product review/etc.
4. Make a decision hopefully.
So many of those pieces of the puzzle just aren't available when you limit your input/output to a system like this. There are obvious tradeoffs here--I just hope we don't lose the ability to make more informed decisions. Already I hate things like Apple obfuscating the full URL in the URL bar.
I think a more likely result is that when confronted with a choice between what people see and what the computer tells them they will believe the computer.
The story is that Amazon sacrifices engineers and others after every big project, success or failure. They have developed a weird human sacrifice culture. Fire phone? It was a failure implementing Bezos's wacky ideas. Engineers had to be purged to atone for it. Fire TV? Huge success but we don't want engineers getting any credit for it. Result: purge some engineers. Echo? Another success that now had to float all those managers from both echo and phone project. Better kick some more engineers off the life raft.
They have a real 'slaughter the goose that layed the golden egg because holy shit! there is a huge expensive golden egg and it's mine all mine, die, die, die..." culture. Bezos probably has an employee purge button built into his desk.
Isn't it more likely that after shipping a 'named' consumer product, many of those engineers rose in value a tremendous amount and rather than negotiate a raise they took a huge salary bump working for Apple/Google?
In my experience, companies are not that great at recognizing and rewarding achievement. So unfortunately, I'll stick to my cynical interpretation.
Also, we're talking about dozens or even hundred of people moving on from Amazon here...
Google and Facebook offer largeish stock grants when you join the company, no ? Google is also known, to offer some kind of bonuses on merit, but that policy is seen as remarkable in tech.
I'd guess Amazon is more political and/or stingy than the above 2.
Anyway I was not referring to big tech specifically.
And Amazon has lots of tiny shell companies like Rawles LLC, fostering distrust and compartmentalization between product teams. Not a healthy work environment.
I remember at one point half of the team (up to director level I mean) knew about a project and the other half didn't. The VP and director had to argue hard to just get everyone in the damn loop. Then they made us sign extra stuff and reminded us that we could be fired or sued if we leaked anything blah blah blah.
When I was at AWS they seemed to be much more open internally than Digital, but the trend was definitely toward secret projects, e.g. WorkDocs and Aurora. Not being micromanaged by Bezos probably makes AWS generally less dysfunctional than Digital.
This is just life at lots of "Heartless" companies. DELL for example, Cuts the bottom X% of staff based on "Performance". Sic, getting a divorce, new children, ask/forced to work odd shifts, working in some managers experimental project...
That is how it works at all. The company took in $107,000,000,000 dollars in 2015 but were just forced to layoff off engineers. Right, no despotic dictatorship banana republic stuff going on there.
It didn't "layoff off [sic]" engineers; they quit.
It could be argued with that much revenue they could afford a better work environment for their employees, but it's not as bad as you make it out to be.
I was looking for a fair comparison between Now, Siri, Cortana and Alexa and I stumbled across this [0]. I find it interesting because it seems you would need all of them to answer all your questions.
Do you have also some personal experince comparing with Now , Siri or Cortana? What went really bad or amazed in terms of speech recognition and speech comprehension?
I can't speak for Cortana, but I have personal experience with Siri, Now, and Alexa. In terms of the actions that I do day-to-day (I would never ask for a dirty joke, for example, but I ask for the weather daily), I would rank them Now, Alexa, and distant third, Siri.
There are two axes that I care about - how many useful commands are available, and how well they can understand me.
Google Now has a pretty wide range of behaviors, but it's the spooky-accurate voice recognition that sets it clearly ahead. In the last year, Google Now has only misunderstood me ONCE.
Alexa also has a very wide range of behaviors (especially with the Skills), but I do occasionally run into things I feel like it should do that it doesn't (yet). To be fair, I also use her the most, so it could be that I push her further than the others. The voice recognition is quite good; I would say it misunderstands me less than 5 times in 100.
Siri is by far the worst. Her abilities are limited, but worse, she has a tendency to funnel commands that she can't do into things she can. For example, asking for the hours of a restaurant will instead show you a Bing search (!) for the restaurant's name, sans hours. This is not useful. It would be better to just inform me that she can't do that yet, file that in a log at Apple, and then add that behavior some time down the road.
She also mishears me 25-45% of the time, AND she seems to be getting worse over time. I assume this is because they are trying to widen her audience beyond a standard California accent, but it's incredibly frustrating for a product to get worse, not better, over time. I now use her only for voice dictation of texts on my iPhone, and use Google Now in app form exclusively for more advanced queries.
> Google Now has a pretty wide range of behaviors, but it's the spooky-accurate voice recognition that sets it clearly ahead. In the last year, Google Now has only misunderstood me ONCE.
> She also mishears me 25-45% of the time, AND she seems to be getting worse over time. I assume this is because they are trying to widen her audience beyond a standard California accent, but it's incredibly frustrating for a product to get worse, not better, over time. I now use her only for voice dictation of texts on my iPhone, and use Google Now in app form exclusively for more advanced queries.
And this is pretty much a huge issue for this market. Entry requires having already spent several billions on voice tech.
You’d have to practically use all of YouTube as training set. Which Google did.
Maybe someday we can have laws that require all neural networks trained with data from others to be in the public domain, even if the people providing training data agreed to the data being used for training.
Then we could compete on other things than "understands voice".
Wait. So Google recognizes a long time ago that machine learning would be important to their business and so invests in a variety of projects to aid with that; Google 411, Google Books, etc. But because Apple either didn't have the foresight or chose not to invest in these technologies, Google should be forced to donate their data sets? To one of the richest companies in the world no less?
> But because Apple either didn't have the foresight or chose not to invest in these technologies, Google should be forced to donate their data sets? To one of the richest companies in the world no less?
No, I’m not arguing for Apple. I don’t care – they can afford it. I, as a student, can’t afford it. You, if you wish to start such a startup, can’t afford it.
And you know what? Google couldn’t do any of this on its own. Instead, they tricked people into working for Google for free.
And now they try to make others pay for the work of volunteers that transcribed YouTube videos, that solved Captchas, etc.
Well, until then, Google's speech recognition API is available (https://cloud.google.com/speech/) so it seems like it should be possible to at least compete at Google levels of SR without even having to understand what training data is.
I don't think it will take billions. LSTM+CTC speech recognition implementations are coming to OSS and I think we have large enough available data sets for training without needing Google size resources. Audio books are a good resource for training and it's not hard to amass hundreds or thousands of hours of training material that way. I think we will have 95%+ accuracy ASR as open source trained networks very soon.
In the case of Amazon Echo/Dot they benefit greatly from the far field array mic for isolating a clean speech signal in noisy/far field environments, would be nice to see generic USB array mics with Linux drivers
Interesting that you rate Google Now as high as you do. I've had somewhat counter experiences.
Specifically, both me and a friend decided to upgrade to Nexus 5X's this past month, he from an iPhone 4S, and I from a Moto G.
Since I'm the one who's been using Android for longer, I've been his go to source for questions about the user experience, woe's about inconsistencies, and investigations on getting things done.
What immediately became apparent is that apparently, the voice integration of Google Now seems quite terrible compared to Apple's Siri. I never use the voice features, but it turns out he is highly dependent on them, using Siri to transcribe most of his casual text interaction, control music, set timers, make quick appointments, and look up facts.
However, when he moved to Google Now he found that while Google Now could in theory do all those same things, that it did so quite inconsistently (failing at minimum 5% of the time, up to 25% of the time). I tried and found the same problems.
For example, I'd unlock my screen, swipe from the lower left to activate Google Now and then say "set an alarm for 10 minutes". Sometimes the little jiggling colored stripes would just never start moving, indicating that it didn't even engage the microphone. Other times it would catch what I say and correctly parse it, even responding "Ok, setting your alarm" and display the alarm card. However, it would never actually set the alarm.
There are plenty of other inconsistencies encountered, this is just one that I experience frequently. My friend can go on for quite a while though about the issues he faces with Google Now.
Anyway, has anyone else experienced issues like this one?
> "Generally, the engineers and product managers at Lab126 quelled their own dissent before it reached Bezos, instead concentrating on giving the boss what they thought he wanted. “We spent so much time trying to anticipate what Jeff would do or say, and read into little words he would say in meetings,” said one former employee. “It would lead to so much additional work.”
This sounds like a recipe for disaster. Very worrying. How have they managed to be so successful with such broken communications?
I am waiting for the day when someone will sneak in those magic words into a TV commercial to wake up Echo, Siri, etc.
Imagine like a DDoS caused by a TV commercial on Super Bowl.
I have the Echo and sometimes it scares the * out of me when something on my TV sounds similar to "Alexa", like "a lexus".
Yeah, they call the activation word a "wake word", but I couldn't help but think of it as a "safe word". It needs to be something that you're only likely to say when you mean it. Which, to me, implies it should be customizable. I should be able to set it to whatever makes sense for me.
This issue comes up in science fiction stories, often set on spacecraft, where there is an AI that is spoken to. On Star Trek, they just said "Computer?" and that got it's attention. But that always seemed a little clunky to me. At first I thought the ship AI should have the same name as the ship, but I quickly realized that probably wouldn't work so well. Picard: "This is the USS Enterprise!" Computer: "Yes?" On Babylon 5, a Minbari character just called his ship, "ship", and in Alien, the ship was Mother, etc. I decided that if I ever wrote an SF story with a shipboard AI (or HAD a ship with a shipboard AI :) I would call it Maru-san.
Such a list doesn't exist for the Echo any more than it does for your cell phone. The Echo is integrated into IFTTT, so any of the actions here [1] are available, and there are also an ever expanding list of actions called 'Skills' [2] that you can add to your particular Echo's list of available integrations (like installing a phone app).
That's all in addition to the basic things it can do out of the box, like set timers and reminders, manage a todo list, play music and books (from an admittedly limited set of sources), tell (mostly bad) jokes, sync with your google calendar and respond to questions about upcoming events, tell you the weather forecast, read off headlines, and a bunch more.
To a greater or lesser extent, it's like a living room installation of Siri.
You could, in theory, list every verb my phone can natively perform. That list is the set of view controllers in the phone's onboard OS apps, plus the set of view controllers in every app in the app store (which is growing, but finite and still easily enumerated.)
The only assumption you need to make is that actions on all server-delivered forms are collapsed into a single "make a network request" verb; otherwise you do get the infinite expanse of actions that is the web.
A list of the APIs would be a good starting point. "Controls your lights" isn't helpful. "Talks to Hue and IFTTT" is.
Just saying it can do anything because internet is misleading and doesn't tell me where the walls are. Unless there already exists a "skill" that can give me load average, player count and tick rate on my Minecraft server.. :)
Maybe I'm old but i still really don't understand the appeal of a device that monitors and datamines all the communication in your home so you can buy products more easily.
Not sure just how seriously this is meant to be taken, but the Echo doesn't monitor anything until you say "Alexa" (and recognition of that is handled in the hardware, locally - if you're worried about devices with net access and microphones, you'll need to keep your phone turned off at home, too). It's fair to assume that once you are interacting with it Amazon mines at least some data, however.
Because the code that understands 'Alexa' is hardcoded into the device. Once it hears that word, that's when audio starts getting sent over the Internet to Amazon's servers.
When is Google going to build a Google Now for the living room to compete with Echo? I want an Echo, but I think a Google product with access to the whole Google ecosystem would be more useful to me.
We know Echo will never play nice with Google (you can't buy Chromecast on Amazon is evidence). If Google built this product, the only feature they probably couldn't match would be buying products on Amazon, which I don't know if I would use anyway.
I got early access to pre-order (can't remember why) at $99 and almost didn't do it. I'm SO glad I did. It gets used all day every day. I grab the last two eggs and say "Alexa, add eggs to the grocery list", hands are full of mess while cooking, "Alexa, set a timer for 5 minutes", etc. The fact that it works with my home automation hub is also nice, when I'm across the room and want the lights off, or want the garage door closed. I could live without it for sure, but it just makes things easier. The voice recognition, especially from far away is incredible. I don't have to yell, or say things exactly right (like in my car).
Knowing what I know now, I'd pay full price- especially with the new functionality and apps that get released every week it seems. I've even written a few of my own.
The speaker itself is just okay, we play music on it all the time for the kids, but there are way better Bluetooth speaker setups, so if that's your primary goal you won't be that excited.
If you can afford it and have a routine of any sort throughout the day where checking the weather, traffic, maintaining a grocery list (especially when you might not be the person going), getting a news briefing or listening to a podcast are part of it, then you'll find it useful immediately. If you don't, to get your money's worth, you'd want to start a routine of sorts.
I have a gen 1 iPad mini taped to my fridge so I have voice control with Siri and I don't think Alexa can do anything Siri can't (for me). I want one, I just don't think I need it. I'd love to automate everything though.
I've had one for the better part of a year, and still can't quite figure out what it's good for, besides kitchen timer and weather.
Most of the questions I ask it are returned with "hmm I'm not sure I understood the question" when the very same question asked of Google Now on my phone returns an intelligent, spoken answer.
Of all the features I wish the Echo had, the best would be.... Google Now integration. Haha.
The full echo happens to be on sale today, and the Echo Dot (which doesn't include the full speaker, only a smaller one, and you can plug it into your own) is about half the price, and does still have the always-on listening mode.
The Echo Dot is ~$90 but my understanding is to buy it you must A) be an Amazon Prime member and B) use the voice interface on an existing Amazon product to place the order (the Echo itself being an obvious example but I guess also the Fire Stick).
I don't remember the specifics, but I was able to buy my Dot without an existing Amazon voice product, they seem to have "fixed" that "problem" though.
I have Prime, so I can't comment to that other issue.
The Echo and Tap are on sale today for 15% off. That brings the Echo to $153.71 or 5 payments of $30.74. The Tap is $111.01 or 5 monthly payments of $22.20.
I just happened to order both yesterday and Amazon refunded the difference without any hesitation.
I've played with the Echo, it is really nice, but ... much like Siri, Cortana and Google Now, using it fully gives up any shred of remaining privacy you had within your own home. I would gladly pay twice as much for an Echo-device that doesn't route all my instructions through $BEHEMOTH servers.
Holy crap that is annoying. Is there a way to pause all animated gifs in chrome? Congratulations, Bloomberg, you captured the essence of a Geocities site. This has to be a not so subtle graphic design joke.
Bloomberg has really gone off the deep end of the design pool. It seemed to start with those somewhat janky, but pretty nice zero-padded headline overlays on the pics, and culminated in this: http://www.bloomberg.com/features/2016-design/ which...well.
I've been using Animation Policy [1] to only play gifs through once and it's worked pretty well so far (in contradiction to the reviews on the chrome store).
Congrats, here's one person who agrees with you. It's like Bloomberg has given some hipster interns full access to design the website. They probably think it's so awesome that they're using all the dynamic components
Annoying gifs seems to be the popular thing at the moment. I've backed several Kickstarter campaigns recently and every single one of them fills their status update posts with stupid gifs.
"An e-commerce company wouldn’t seem like an obvious leader in [augmented reality]."
Amazing that even today articles are still calling Amazon an e-commerce company. I wonder if this is an image Amazon actively cultivates or if it's just consistently lazy reporting?
It's possible the author of the article knows that Amazon is now a tech company as well, but is assuming that their audience does not yet realize this. My impression is that people who don't work in tech don't know this; I have explained many times to people outside of our world how Amazon is one of the most important tech companies in the world, and that discussion always starts with a puzzled look from them.
Yeah, that's kind of my take as well. But I put that in the lazy reporting category.
Seems like it would be worth the paragraph explaining the misconception and helping people realize why the seemingly weird stuff Amazon kinda does makes sense.
Don't feel bad -- we at Bloomberg always get classed a "media" company when the media properties came about a decade after the software (& hardware). It's just easier to say because it fits better into public expectation. To most people Amazon is e-commerce because nearly everyone buys things from them online or via mobile apps.
If this sub-thread were not about being specific in one's characterization, I would not point this out, but:
> online or via mobile apps
As far as I'm concerned 'mobile apps' operate 'online' and it isn't clear the distinction you are making. Is it common to refer to 'online' as via a web interface (ports 80/443)?
I think that's the distinction and I think it's a valid one. My experience on Aliexpress, Amazon, and Ebay (to name the 3 I use the most) is that the web and mobile interfaces and use cases are still quite distinct from each other. On Aliexpress, I sometimes use the mobile to checkout for the explicit discount, but never search from mobile. On Ebay, I respond to bidding alerts on mobile, but don't initiate transactions/searches on mobile. On Amazon, I basically never use their mobile app (and usually use google to search Amazon's [web]site).
How would you classify them? (genuinely curious) Everything they offer centers around consumerism, and separating your money from your wallet in exchange for some product or service. Even things that don't quite fit into that category (the tech stuff) are ultimately there to facilitate the purchasing of said goods and services.
Not trying to be pedantic, but isn't every business trying to separate "your money from your wallet in exchange for some product or service?" The only potential difference I could see (looking at it through this lens) is whether it's a consumer or business oriented company.
For what it's worth, I'd categorize Amazon as (principally) an eCommerce company.
I suppose that is so. I meant more in the sense that all of the tech they develop makes it easier for consumers to consume other products and services. Their tech isn't an "end game". I.e. They're not making things like smart fridges, rc cars, and etc. (Yet - unless that smart fridge allows you to order the products they sell and watch their streaming video in the kitchen.) All (maybe most?) of their tech products facilitate the purchase of the products made by other companies, whether it be allowing the consumer to purchase things easier, or enhancing their fulfillment process to get the products to the consumer faster. I guess I was more debating the fact that some like to classify them as a predominantly tech company, but I neglected to actually say that.
> Generally, the engineers and product managers at Lab126 quelled their own dissent before it reached Bezos, instead concentrating on giving the boss what they thought he wanted. “We spent so much time trying to anticipate what Jeff would do or say, and read into little words he would say in meetings,” said one former employee. “It would lead to so much additional work.”
A moment passes.
"Dad, Alexa just told us...!"
My geek dad game is strong.