ARCHITECHT Daily: Apple HomePod could be a defining moment for consumer AI

Just a heads up that I'll be in Seattle part of this week at the GeekWire Cloud Tech Summit. This wil
ARCHITECHT
ARCHITECHT Daily: Apple HomePod could be a defining moment for consumer AI
By ARCHITECHT • Issue #90
Just a heads up that I’ll be in Seattle part of this week at the GeekWire Cloud Tech Summit. This will almost certainly affect the time you receive these emails Tuesday through Thursday.
Apple announced its HomePod smart speaker / home assistant on Monday, so all parties in the battle to own the smart home are now present and accounted for. And because Apple is making its entrance in typical Apple fashion (i.e., expensive—$349—and aesthetically pleasing), HomePod’s success relative to the Amazon Echo and Google Home devices could teach us a whole lot about what consumers really value when it comes to artificial intelligence. (Also digital platforms.)
I’m gonna resist the urge to do too deep an analysis right now, because HomePod doesn’t come out until December and I assume there will be lots of activity between now and then. Apple, Amazon and Google should all be racing to improve their speech recognition models, add new features, bring on new partners and (I assume in the case of Amazon, at least) adjust their pricing. You can rest assured the Amazon-Sonos integration will be complete by then, if Amazon doesn’t just buy Sonos outright in the name of PR and greater market penetration.
Aside from the market dynamics, however, the introduction of HomePod could also illustrate just how much consumers care (or don’t care) about superior AI technology. If we were to break things down as they stand right now, I think it would look roughly like this:
  • Apple: Expensive product; dominant mobile platform; good, if limited, AI assistant with Siri; lots of data around music/media content and voice queries.
  • Amazon: Relatively inexpensive products; first-mover advantage; good, if limited, AI assistant with Alexa; lots of data on shopping and voice queries, and some on media content.
  • Google: Moderately priced product; broad mobile platform; superior AI all around; lots of data on everything but shopping.
  • Microsoft: Microsoft has good AI, but is not selling its own home-automation products yet, and I don’t suspect “Cortana-powered” will prove too big a draw to third-party products.
Before HomePod came along, Amazon and Google were battling have the best the consumer-developer-cloud flywheel, where more products sold means more developers means more money spent on cloud infrastructure. To some degree, it was a fight between a ruthless competitor willing to accept thin margins in the name of moving units, and the company now synonymous with cutting-edge AI. In some circles, both companies have reputations as being particularly greedy for consumer data. 
Apple and HomePod have access to less data and, presumably, bringer a narrower range of knowledge than does Google Home. HomePod presently costs twice as much as the full-size Amazon Echo, and several times as much as the Echo Dot. However, HomePod brings with it Apple’s dominant iOS platform, reputation for stellar product design, and a strong focus on playing music (a space where Apple currently dominates Amazon and Google).
If HomePod catches on like the iPod, iPhone and iPad before it, we could take that as a referendum on the importance having the best models in the smart home: If a product is otherwise superior or masters the actual use case people care about (or maybe just is made by Apple), good enough is good enough. The silver lining for Amazon and Google is that their clouds will likely be hosting the backends for many third-party HomePod apps.
On the other hand, if HomePod struggles because Siri isn’t living up to her end of the bargain, Amazon and Google will never let consumers, or Apple, forget it. And they’ll be laughing all the way to the bank as their flywheels remain spinning at full speed.
For what it’s worth, I have several Sonos speakers, one Echo, one Echo Dot, and an Echo Show on pre-order. I made my bet with Sonos and then the original Echo before Google Home was announced (long before, in the case of Sonos), and assuming the Echo-Sonos integration works nicely, I plan stick with it because wireless music is a big deal in my house and my money is spent. 
However, I do sometimes yearn for the superior voice-search capabilities of Google Home, and while I’m not a big iOS fan, I do respect Apple products. I don’t know what it would take to woo me away from Amazon, but I’ll know it when I see it.

Sponsor: Cloudera
Artificial intelligence
Teams had to build  agents that learned to collaborate while playing Minecraft. I don’t know a whole lot about Minecraft, but it seems like it could be a pretty good environment for training models that adapt more easily to the real world.
It’s easy to get excited about GPUs and TPUs, but there’s a large contingent of folks who think FPGAs can still play a big role—including, apparently, at Harvard’s engineering school, where they’re building an integrated circuit using FPGAs.
There’s not a lot of new info here, to be honest, but it is worth thinking about how we talk about the higher-level applications of deep learning and other techniques. Are they AI? Cognitive computing? Something else?
This focuses on some scary-looking faces made by the pix2pix project, generated from sketches of faces. But as GAN techniques get better, they’ll be producing higher-quality synthetic data and making models much better.
If this author’s experiences—that humans want to develop “relationships” with AI assistants, and will pretty much to what they suggest—ring true across the board, then we really do have some thinking to do as AI systems get even better. 
This time, via a technique called latent attention networks. If deep learning is going to be deployed in certain fields, it’s going to have to work and be auditable or understandable at some level.
arxiv.org  •  Share
Sponsor: DigitalOcean
Cloud and infrastructure
This research could be super-important not just to the adoption of smart devices from wearable to cars, but also to IBM. The next-generation chip competition is still anybody’s game, and IBM needs a win. 
Also, if you prefer your advances in chip technology more theoretical and less silicon-based, check out this carbon-based approach from University of Texas at Dallas researchers.
If you’re into predicting where cloud infrastructure (or, more accurately, cloud platforms) is headed, take note of the outliers here.
www.cncf.io  •  Share
Aardvark and Repokid are tools for gaining more control over identity and access management at large scale, building on top of available AWS tools for that task.
medium.com  •  Share
This is a glimpse into one of the areas where large web companies don’t have complete control. They still depend on relatively predictable supplies and prices on things like memory, and apparently they’d like to buy from a manufacturer they know.
Facebook keeps putting on its @Scale events, each of which focus on different aspects of running a large web company. Here are presentations from Facebook, Amazon, Microsoft, Spotify and more about optimizing software lifecycles, more or less.
It’s hard to argue with this premise, although it’s also hard to predict exactly when this new future will take hold. CIOs are thinking about both these technologies, but really are still moving toward cloud computing and big data, in general.
Media partner: GeekWire
All things data
 I suspect the actual number somewhere in the ballpark of Hortonworks and Cloudera revenue; probably less but certainly not higher. If it’s higher or soon to be higher, that will make MapR’s eventual IPO much more exciting.
I always praise innovative ways of collecting important, but hard to gather data. You’d like to think cities and other institutions can use stuff like this to improve air quality and quality of life, overall.
This is a good analysis, but is more focused on cloud than on cloud-native. With regard to the latter, things like Kubernetes+Helm (part of Microsoft’s Deis acquisition) and DC/OS can simplify the process of deploying and managing data infrastructure.
redmonk.com  •  Share
Its five partners are Intel, Qualcomm, Pacific Northwest National Laboratory, Georgia Tech and Northrop Grumman. Analyzing networks is possibly more important in the defense world than in other places.
Did you enjoy this issue?
ARCHITECHT
The most interesting news, analysis, blog posts and research in cloud computing, artificial intelligence and software engineering. Delivered daily to your inbox. Curated by Derrick Harris. Check out the Architecht site at https://architecht.io
Carefully curated by ARCHITECHT with Revue. If you were forwarded this newsletter and you like it, you can subscribe here. If you don't want these updates anymore, please unsubscribe here.