Hacker News new | past | comments | ask | show | jobs | submit login
Show HN: Siri, open the garage (delian.io)
129 points by MIT_Hacker on April 6, 2014 | hide | past | favorite | 44 comments



This post got nailed by the voting ring detector, but I'm restoring it because we want to see original work on HN.

A "voting ring" is when people get friends to upvote their stuff. This is against the rules. We want stories to be on HN because they're good, not because they were promoted.

It's sadly common for a great Show HN post to get demoted because its creators, eager to get it on the front page, tried to game it. I've noticed a pattern, too: usually their gaming technique is pretty weak. Perhaps that's because they're creators, not promoters. Unfortunately, it has the side-effect of making it certain that the ring detector will nail their otherwise good post, while we carry on the usual cat-and-mouse game with people pushing crap.

I've got what I believe will be a sweet solution to this problem, but it awaits time for implementation.

Please everybody, don't ring-vote your posts; just take your chances with HN's randomness. If a post is solid and hasn't gotten any attention yet, a couple of reposts is ok. Be careful not to abuse that, though, since we penalize accounts for reposting too much.

I'm going to demote this comment as off-topic so it won't get in the way of the real discussion. Send any moderation questions to hn@ycombinator.com.


Which do you think is more likely:

- I know! We'll set up a voting ring to game my blog post about a garage door opener onto the front page of Hacker News, even though I don't get any kind of benefit aside from exposure to my cool project!

- Hey $socialnetwork friends, check it out, I submitted last weekend's project that you helped me on to Hacker News!, followed by folks finding the link there upvoting it because it's cool.

You, and pg before you, apparently default to option 1. Even in the way you've worded this comment you are implying malicious intent, as if the author of the blog post tried to game the article onto the front page. Honestly, unless there's a tangible benefit to the traffic such as advertising or leads, there's no reason to game HN and most of the "voting ring" stuff you are likely observing is organic-ish, not intentionally gamed upvotes from social sharing.

Option 2 is something I have personally observed dozens of times and you should account for it in your thinking and code.


Um, actually I believe that dang's post is him literally saying that cases of Option 2 are mistakenly classified as Option 1, and that this is unfortunate. And that people should know about the voting ring detector so they don't accidentally trigger it. Oh and that in his spare time he is working on a better detector.

Having a voting ring detector is a good thing. It is no doubt hard to tell organic friendly votes apart from sophisticated voting ring bots, and lots of good posts probably get flagged for this. But without such a detector, there would be a programmatic way to ensure posts get to the top of HN, which defeats the whole voting system.


Neither we nor our software can tell much (if anything) about intent from the data, but I certainly don't think it's always malicious—I know for a fact that it's not.

The ring detection algorithms are vital to the quality of HN. I've looked at a tremendous amount of data on this, and I don't think it supports your interpretation. But certainly the algorithms and moderators both make mistakes. As I mentioned above, I think I may have a way to compensate for that, but I don't know when we'll be able to implement it.


Option 2 is literally a voting ring. Why else would somebody tweet about the empty HN discussion thread they've created?


Option 2 shouldn't be a link to an empty HN post. It should be a link to the content. If the people who read it then decide it should be submitted to HN/upvoted they can go and do that.


> I then jokingly opened siri and said “Open the garage”. Siri responded with “I can’t find an app named ‘The Garage’”. Lightbulbs went off. We realized we could get around the fact that Siri has no API by making apps named the phrases we’d like to recognize.

Priceless!

I wish Siri had an API though.


I find it curious that both Google and Apple have held back on this. It seems so obvious that there's an entire platform to open up here based on this kind of thing, but neither one seems to be interested in doing it. Instead they are privately noodling away, just incrementally building up their own vocabularies which are really pitifully small in the wider context of things. They must both know there would instantly be an explosion in 3rd party support the minute they add useful APIs, and it would be a major selling point for either OS ... I wonder why they are both so reluctant to do it?


Google Now is integrating some information extracted from the user inbox, when the received e-mail includes the right schemas:

https://developers.google.com/gmail/actions/

By the moment it requires registration, but we are using it and works nicely. It's one step in the right direction, I suppose.


I feel like this has to come with iOS 8 for Apple to stay competitive against Google Now.


Also to compete with Cortana on windows phone, which lets you register voice commands to run specific apps and provide it input [1].

[1] http://channel9.msdn.com/Events/Build/2014/2-530 - Integrating Your App into the Windows Phone Speech Experience


In fact apps can register for voice commands since the release of WP8. Microsoft just improved it so you don't have to register e.g. 1000 movie names to enable the WP voice recognition to understand them, but telling WP to listen for a movie name.


But I thought Google Now doesn't offer an API either? I use both Siri and Google Now (I have an iPhone and Nexus I find myself regularly switching between) and I feel like they are two different products, not meant to compete (yet). Yes, Google now can answer questions like Siri but because it's main focus is on showing me stuff I need when I need it I tend to forget about the Q&A part of the product. I also feel that's Siri's Q&A is better. Google Now needs to pull in more sources and open an API to keep competitive with Siri. Siri needs an API and more contextual awareness to keep competitive with Google Now.



Windows Phone user here: we got this API for 1.5 years and Cortana just brought it to a higher level :)


I did this a few years ago using text messages and Twilio. Not quite as slick as "Siri, open the garage", but it worked! http://rumsey.org/blog/2011/11/voice-controlled-garage-door/

"Siri, text house open the garage" was as close as I got.


You know, I don't hate that as a user experience. It kinda turns the house into an "intelligent" entity. "Siri, text house, I'll be home in 5 minutes" and it would turn on the lights five minutes from now or whatever you like.


That duct tape... and electrical stuff hanging off the ceiling... makes me cringe! EEs in my dad's company used to call that kind of project an "ikebana." :-)

http://en.wikipedia.org/wiki/Ikebana


Our housemates seem to have decided that it's just over the border of the tradeoff between convenience and fear :)


Not just hanging from the ceiling... hanging from communication wires on the ceiling. :shudder:

Still, some sins are excusable in the name of proof-of-concept. :)


So it's art then?


A team at PennApps had a project [1] which used a DNS-override to take advantage of the fact that Siri defaults to Googling things when it doesn't have a native approach. They could then parse these requests and see if they could do something cool without before going ahead and just Googling it. The example they used in the demo was controlling a Phillips Hue light with Siri commands. I thought this was awesome, one of the better hackathon demos I've ever seen. There really should be an easier way to do this an spoofing Google DNS, which I'm not comfortable with even though I trust these guys.

1 - http://betterthansiri.com/


Very cool hack! I would never have thought about implementing each command ("open the garage") with a specifically named app ("the garage"). I guess it'll work until you already have a (real) app called like the thing you want to open :)

An easiest way would have been to use Wit.AI API [1](disclaimer: Wit team member here). You can create any command in minutes and use any device capable to stream audio to the API (including your iPhone) as client. I'd be interested to have your feedback if you have time to check it out!

[1] https://wit.ai/blog/2014/02/12/speech-api


There is SiriProxy[1] which lets you respond to arbitrary commands by proxying requests to the Siri backend. Apparently doesn't work with iOS 7 yet unfortunately.

[1] https://github.com/plamoni/SiriProxy


This is great, although it's too bad it requires such a clever twist to integrate with Siri. Cortana should shine in situations like these: http://9to5mac.com/2014/04/02/microsofts-cortana-siri-clone-...


BAD ASS!

I see in the near future a folder named "commands," on your iPhone filled with long name apps just for automation. Are you using https or some type of SSH keys to authenticate?

I know you said your running wired instead of wireless, but it sounds like you have to run some type of lightweight server to listen to the http requests. I guess you could just port forward/trigger internally through the router.


Thanks man!

Currently it's security through obscurity... so not great. I'll make the arduino only accept requests with a token next weekend.

Yeah if you take a look at the arduino code on github it's a super light http server. Currently I just do port forwarding to access it.


Of course even if you require a token, it doesn't stop replay attacks.


Notice that saying "Siri, open the pod bay doors" doesn't work.


An example using a Raspberry Pi and SiriProxy, which actually handles more than just the "open" command from a year or so back: http://www.raspberrypi.org/forums/viewtopic.php?f=63&t=25118.


I've looked into playing around with Siri for home automation stuff also but unfortunately, SiriProxy ( https://github.com/plamoni/SiriProxy ) doesn't work with iOS 7 yet ( https://github.com/plamoni/SiriProxy/issues/542 ).


Things like that are perfectly possible without any internet-connected services with open source software like CMUSphinx. And as an addition you can get better features like continuous listening (no need to push any button), flexible vocabulary and even speaker identification (only you will be able to open the garage). I really wonder what is cool in Siri doing that.


Nice hack! Could you elaborate on why arduino + ethernet shield will be a better choice than raspberry pi + wifi? I'm genuinely curious because I've been looking at both for a hardware project and would love to know the advantages (and gotchas) of both approaches. :-)


Consider Electric Imp too - very small, very cheap, wifi built in, and the web connectivity is handled for you so you don't need to set up/secure/manage an http server.

I've got one that will control my electric desk from my phone, one day soon.


The Imp seems cool, but it's price point (when combined with the breakout board) is higher than a Raspberry Pi and Usb Wifi dongle. Not only is the price higher, the hardware isn't nearly as capable.


Entirely familiarity. I had used Restuino and Arduino a lot before :)

I wanted to do this as quickly as possible rather than playing around with Linux and learning a new framework!


Very cool!

I use a hack to have Siri retrieve my bank balance via a SMS. This comes in handy while in the car and you want to balance your checkbook in your head.


Very cool.

But when you say "app" what exactly does that mean? Can it just be a web app that you save a shortcut to? Or is this actually a native app?


In order to take advantage of the Siri "api", it has to be a native app. All the native app does is open a webview with a specific link.

Siri doesn't recognize the names of bookmarks you save to your home page.


This is very cool, but I'm surprised a garage opener with this type of interaction doesn't already exists.


I mean, garage door remotes cost $5, the battery lasts for years, and all you do is press one button. I just can't see why you would want to make a voice-activated garage door remote.

(Integrating it into a phone is slightly more compelling)


In the case of this house specifically, most people walk/bike rather than drive, and walking around with a garage opener in your pocket isn't as nice as talking into your headphones.


Why would someone walking need to open the garage?


the door is locked and we don't carry keys :)




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: