Menu

Colin Devroe

Reverse Engineer. Blogger.

Deep Learning and Siri’s voice

Apple:

The training speech data contains a minimum of 15 hours of high-quality speech recordings sampled at 48 kHz. We segmented the speech into half-phones using forced alignment, i.e., automatic speech recognition to align the input phone sequence with acoustic features extracted from the speech signal. This segmentation process results in around 1–2 million half-phone units, depending on the amount of recorded speech.

The entire methodology sounds very impressive.

While I’ve done some basic research on machine and deep learning in the recent past, it wasn’t nearly enough to keep up with this entire article. This is something I do often. I read an article and if there are large portions I simply cannot comprehend I’ll do research until I grok it. To that end, I plan to set aside some time in September to do enough research on DL and ML for me to understand posts like this at a basic level.

WWDC 2017 wish list

It has been an exciting year for developers so far. Facebook is making the camera a platform, Microsoft is making cloud computation happen with two clicks of a mouse, and Google is doing everything that everyone else is doing plus a billion more things.

WWDC is next week. So what are my wishes? Since I use Apple products far more than Facebook, Microsoft, or Google products, I tend to want more specific things from WWDC.

Here is my list, in no particular order:

  • Shared iCloud Photo Libraries. There are a few ways that Apple could do this. The simplest, in my opinion, is for me to allow access to my iCloud Photo Library to anyone in my iCloud Family Plan. All photos taken from all family members in the same library. Perhaps somehow filtered by device or person. That’s it.
  • iCloud data Family sharing. I buy 1TB for me and 50GB for Eliza. I’d like to purchase 1 set of storage for both of us and be able to share the space allotment. Not only to save the $1 per month but also to combine the accounts.
  • An all-new Mac App Store. The app store app on macOS feels incredibly dated and fragile. It doesn’t seem to work nearly as well as its iOS counterpart. This part of the OS should be rock solid and perform very well. But there are little idiosyncrasies (like how the progress bars look weird when downloading, or how the fonts look…).
  • App Store demos. If Schiller is serious about bolstering the App Stores I think it is time to bring true demos to both App Stores. I don’t have a silver bullet model (7 days, 14 days, 30 days, etc.) but I do believe this is achievable and would be a boon for app developers.
  • App slimming. Apple announced something about this a few WWDCs ago I think. And I believe it is on the developers of the apps to make their apps as small as possible. However, I think Apple can lend a hand to the most popular apps (Facebook, Uber, Twitter, Messenger, SnapChat, Instagram, etc.) and ask that they somehow combine frameworks or resources in order to shrink their apps considerably. Just think, every megabyte Apple can help slim from those apps would be hundreds of terabytes of data usage saved.
  • More Camera app filters. I use the Camera app a lot. Even with the numerous camera apps available (which I’ve likely bought dozens of and installed even more of over the last decade of having an iPhone) I usually gravitate to the camera that I can swipe to from my Lock Screen. Once and a while I’ll use a filter. Either in editing or even in shooting. Its fun. I’d like to see more added. Like, 50 more with me being able to select my top 12 somehow.
  • Apple Prime. Amazon Prime comes with a slew of things; music, movies, photo storage, free shipping, etc. etc. I’d like a one-payment Apple bundle that would give me iCloud space, iCloud Photo Library, Apple Music, Apple TV (or whatever their TV service will be if it ever happens), etc. Some “prime” membership per year that I can be all-in on Apple stuff.
  • Rename iCloud Photo Library? Perhaps it should just be called Apple Photos. If every app simply had an iCloud switch that allowed you to store / sync its data with iCloud we wouldn’t need separate names for things. The apps simply can have an iCloud feature.
  • Apple Watch Series 3. I’ve said since the beginning that the first Apple Watch I’d buy would likely be the Series 3. (Actually, I begged Apple not to make a watch at all. But, since they did, the first one I’d likely buy might be the next version.) My wish list for series 3 would be slimmer (less tall), no phone needed at all for it to function (network-connected with no additional plan), much, much faster, and easier to update/install apps. Essentially, a stand-alone device rather than an iPhone accessory.
  • Apple Photos improvements. Only a few of my wishes from May 2016 have been addressed. I would like to see Apple Photos get substantially better this year. However, Apple seems to improve things much slower than they used to. (Remember the iPad 1 to iPad 2 jump? I wish we saw more of that speed from Apple)
  • More iOS Extensions. Perhaps Apple’s purchase of Workflow won’t bear fruit so quickly, but I’d like a lot more Extensions in iOS. I often find myself limited in what I can do with a file on iOS. I think it should be much more powerful to send files from app-to-app or to a service or run a routine on it, etc. I think we’ll see that in future iOS releases.
  • Siri. I’ve nearly given up on Siri. My WWDC 2016 wish list still has Siri items on it that haven’t been addressed. Still, if Siri was 5% better I’d take it.
  • Apple Maps accuracy updates. Apple Maps has improved a lot since its debut. Its design is far better than it was and its feature set has grown too. But, for me, its accuracy is still terrible. Google Maps gets me to the correct location every single time. I can’t remember when it hasn’t. Apple Maps routinely gives me the wrong location when I ask Siri for directions somewhere. Somewhere around 75% of the time. Three out of four. This is not an exaggeration. So, I do not use it. With each update to iOS I give it another try. Then I go back to Google Maps. I don’t think Apple Maps needs any new design, or any new features, it just needs to be accurate. Side note: I was in Philadelphia with Eliza recently and we relied on Google Maps for all transportation. It was excellent at getting us around via Uber, walking, and driving. It was perfect the entire time.

I’ll stay away from any hardware wishes as I don’t have any needs currently. I’m all set on the hardware front. Our iPhones, iPads, MacBook Pro, and iMac are all just fine the way they currently are. And, I don’t need an Apple Home (if they release one) because I have enough terrible Siri devices laying around the house.

Google Assistant for iOS is a clunky mess

Karissa Bell for Mashable:

For starters, Assistant’s iOS app is a confusing, disjointed, mess. You’d think the Assistant would be able to easily link up with all your other Google services, but that wasn’t the case in my initial testing.

It isn’t often that I agree with an article on Mashable. However, in this case I very much agree.

Using Assistant for iOS for a bit yesterday I found myself scratching my head thinking “but, I thought I could do this, or that, or that?”. In reality, the Assistant on iOS as it stands is Google’s search with voice input on top of it. Just like Siri. Which is equally frustrating to use.

One quick note here: Google does not have the access to iOS APIs that Apple does. For this reason Assistant is neutered from the jump. However, Google does a bad job explaining that and so user expectation is pretty high when I first installed the app. Assistant on iOS will likely never be as good as the Android or Home Assistant and that should be clear.

One other quick note: I think the entire tech industry began using the term “AI” a decade too soon. “Bot” is ok. A bot can respond to a set of commands and only those commands. That feels much more accurate when describing Google’s Assistant. But calling features like Cortana, Siri, Assistant, Bixby (or whatever Samsung’s assistant is) is a huge, huge stretch of the term AI in their current states. And likely will be for a number of years to come.

Hey, umm, Siri?

I was happy this week to see that the topic of how far behind Siri is came up on many tech blogs. It is a topic I’ve thought, but not wrote, a lot about. In 2012 Siri was ahead on ability, but behind on speed. Earlier in 2016, prior to WWDC, I wrote a WWDC wish list and in it I wrote that I had hoped:

I hope Siri can do a lot more – I think we’re overdue on being able to say things like “Hey Siri, send the photo I just took to my wife.” Or “Hey Siri, open Spotify and play Jack White’s Blunderbuss.” Or “Hey Siri, find a note in Simplenote that I created on May 15th”. Or, even, “Hey Siri, show me all of the photos I’ve taken in Hawaii.”

Based on what I saw at WWDC I thought I was going to get some of these things. But I was wrong. There are a few things that Siri has improved in iOS 10 but overall it seems that it is falling further and further behind with every announcement from Apple’s competitors.

Even without any competition from other companies I still think Siri should be better than it is. The examples I gave above should already be possible. I’ve had some notes regarding a few queries that I thought Siri should be able to handle by now but she cannot and so I’d like to take one of those and add them to my wish list.

“Hey Siri, how long until it rains?”

This query would be huge for me personally. I use and open Dark Sky at least twice a day. Because I hike and kayak and go for a jog outdoors I like to know, as accurately as possible, when it will start raining. Do I have 30 minutes to get in a quick jog? Or do I have a few hours to go kayaking? I’d like to know and it’d be convenient for me if Siri could tell me.

Walt Mossberg:

For me, at least, and for many people I know, it’s been years. Siri’s huge promise has been shrunk to just making voice calls and sending messages to contacts, and maybe getting the weather, using voice commands.

Why are people only using it for these things? John Gruber explains:

The big problem Apple faces with Siri is that when people encounter these problems, they stop trying. It feels like you’re wasting your time, and makes you feel silly or even foolish for having tried.

Even if Siri has gotten better no one will know because they’ve already given up. I know in a lot of cases I have. And that may be a harder hurdle for Apple to jump over than simply improving Siri’s capabilities.