ArchiTECHt Weekly: AWS makes its move to the edge; AI showdown: China vs. Canada?

Say what you will about Amazon Web Services, but the company is not afraid to take some risks. (Perha
ArchiTECHt Weekly: AWS makes its move to the edge; AI showdown: China vs. Canada?
By ARCHITECHT • Issue #1
Say what you will about Amazon Web Services, but the company is not afraid to take some risks. (Perhaps when you own a market like AWS does, it becomes a little easier to throw caution to the wind.) Case in point: the IoT Button Enterprise Program AWS announced on Thursday, which stretches the definition of cloud to its edge. Or maybe just to the edge.
Essentially, AWS is selling enterprises buttons that look a lot the Amazon Dash button, only with custom branding. And instead of ordering Tide or milk or whatever, a press of the button triggers whatever function the company has programmed it to do. That could be a back-office task, or a consumer-facing service.
Here are the examples AWS provides of how one might use the new buttons:
Reordering services or custom products such as pizza or medical supplies
Requesting a callback from a customer service agent
Retail operations such as a call for assistance button in stores or restaurants
Inventory systems for capturing products amounts for inventory
Healthcare applications such as alert or notification systems for the disabled or elderly
Interface with Smart Home systems to turn devices on and off such as turning off outside lights or opening the garage door
Guest check-in/check-out systems
The IoT button idea was audacious to begin with, and even more so when it’s targeted at enterprise customers. It’s also brilliant—just like the idea to start AWS in the first place when the whole world thought of Amazon as a bookseller. The buttons, along with the AWS Greengrass service the company announced at re:Invent in December, are early efforts to own the edge while everyone else is still fighting to get a piece of the centralized cloud (while, of course, still making money on the cloud backend).
If the Internet of Things really is the future, that is a great position to be in.

The IoT Button. Source: Amazon Web Services
The IoT Button. Source: Amazon Web Services
Big $$ pour into Chinese, Canadian AI efforts
Most Americans might think of companies like Google and Facebook when they think about who’s leading the charge in artificial intelligence, but across the pond from Silicon Valley, Chinese companies are not lying down. Chinese search (and general web) giant Baidu made two big splashes on Monday by announcing its new augmented reality lab, as well as the hiring of former Microsoft SVP Qi Lu as Baidu President and COO. According to a quote from Lu in Baidu’s press release, AI will be a major focus of his new role:
“Baidu is well known as one of China’s top technology companies, and is already recognized on the global stage as a leader in AI. I am excited to help realize Baidu’s visionary AI strategy. To be part of Baidu’s evolution into a world-class technology company for the AI-era is a tremendous opportunity.”
MIT Technology Review published a short post detailing some of the other big AI investments Chinese companies have made recently.
I’ll also draw a connection to Alibaba’s $800 million sponsorship deal for the Olympics, which Bloomberg reports is in part to “showcase its nascent cloud computing business on an international stage.” In part because it’s more evidence that Chinese companies do not want to lose their customers to American competition and unload the cash truck in order to prevent it, but also because I expect Alibaba’s cloud story will have an AI element, as well.
These types of figures make Microsoft’s $7 million investment in Montreal AI research institutions look small. But on the other hand, the company did just acquire a deep learning startup from down the road in Waterloo last week. The draw of Montreal, of course, is less about its economic opportunity and more about the network of AI researchers and entrepreneurs hanging around, thanks in large part to deep learning pioneer (and new Microsoft adviser) Yoshua Bengio.
WIRED has a good profile of Maluuba, the company Microsoft acquired, as well as the type of natural-language research it was doing and why Microsoft is so excited about it.
Qi Lu: Source: Microsoft
Qi Lu: Source: Microsoft
This week on ArchiTECHt
The economics of private clouds, the rise of ‘serverless’ and the business case for carjacking
In this podcast, Joyent CTO Bryan Cantrill discusses why Samsung bought his company, why serverless computing is catching fire, and how Amazon Web Services is overtaking the IT industry. Continue reading on ArchiTECHt »
Stitch Fix CTO talks data science, modern architectures and moving up the engineering ladder
Cathy Polinsky shares her thoughts on what works and what doesn’t when it comes to building software and teams. Continue reading on ArchiTECHt »
Listen to the ArchiTECHt Show podcast
What's news in AI
Derrick says: Conventional wisdom would say Google still has more audio data via YouTube and smartphones, but Amazon probably does have more, better data on how people interact with home digital assistants.
Derrick says: This is potentially very valuable for researchers. And if Jeff Dean and Yoshua Bengio are excited, then I’m excited.
Derrick says: AI has certainly been a the buzzword du jour for security in general lately, so why not for payment fraud specifically? PayPal is already investing heavily into AI research in this space, so it makes sense that someone should try bringing it to a broader audience with less R&D and less data.
Derrick says: This is an interesting theory, but it might be limited in scope. For many companies and developers not too interested in squeezing out that bit of performance or innovating on AI, everything is becoming a whole lot simpler thanks to cloud services and really good OSS.
Derrick says: Kudos on open sourcing it. But can anything attract attention away from TensorFlow and the other big boys’ AI frameworks.
Derrick says: One has to wonder how much sway IBM will have in whatever AI “ethics” are ultimately upon. Watson was an amazing ambassador for AI several years ago, but now it’s companies like Google and Facebook that are driving the discussion and innovation. (Although IBM can still make a mint via AI if it can master commercialization for large enterprises.)
Derrick says: Yes, Neurala CEO Max Versace is actually a member of the Versace family. More importantly, the company, which has been around since 2013, raised a $14 million Series A round and wants to help devices run machine learning algorithms locally using standard CPU and GPU architectures.
What's news in the cloud
Derrick says: Oracle cannot win against any of those providers in a contest over scale or price or open source or, frankly, public perception.  The second a customer decides it wants to re-architect its application using open source technologies—or that cost, global scale, ecosystem or lineup of services is the most important criterion—it’s hard to envision Oracle as a truly viable option.
Derrick says: Interesting take, but I disagree. Microsoft needs to support what customers want to run, and they want to run Docker containers. But don’t think for a second there are too many choosing Docker tools over Kubernetes for orchestrating them.
Derrick says: Backblaze made its name several years for open sourcing the design of the massive-scale storage “pods” that underpin its cheap and unlimited consumer backup service. Since then, it has been moving into the enterprise—first with an object storage service in 2015 (which it calls the cheapest in the cloud) and now with $50-per-year enterprise PC backup.
Derrick says: That’s up 18 percent, while non-cloud gear revenue is expected to decline 3 percent. Notably, private cloud spending is on the rise, too. If you’re a company used to fat margins on hardware sales, these are not good trends. If you’re a company supplying components or assembling gear for cloud, or selling white boxes, this is music to your ears.  •  Share
Derrick says: For the same reasons that all open source technology matters: it’s the answer to the lock-in inherent when using cloud provider services. Whether it’s a serverless framework or Apache Spark, having the option of portability is a good thing.
Derrick says: Docker is a force to be reckoned with, even for the container orchestration vendors and cloud providers looking to pigeonhole it at the file format layer. Every improvement to the core Docker technology is another reason for developers to trust in the company as they move up the container stack.
Derrick says: This clearly has nothing to do with cloud computing, but I really want to draw a connection between BirchBox retail stores and companies moving from the cloud to their own data centers. Also, this news supports my theory that there are some thing people just want to buy in a store.
What's news in (big) data
Derrick says: This is one of those surveys that’s enlightening for reasons not directly related to its actual content. Specifically, it explains why we’ve seen such minimal investment in “enterprise” AI software and so much more on job-specific APIs and consumer devices. If big data efforts are still just ramping up, don’t expect many enterprises to be training their own deep learning models any time soon.
Derrick says: Microsoft Research is working on a method to protect people from being tracked if they have location services turned on. This is one of the many conundrums of big data, where sharing more data arguably results in a better service but a burning suspicion you’re being watched.
Derrick says: This is bad news for the companies selling distributions of these technologies, some of which are eyeing IPOs—even if their tech is secure. You have to hope these were non-essential data stores, which, let’s be honest, almost has to be the case. But, still, it is not a good question to have to answer as you’re selling to security-conscious enterprises.
Sagely advice
Derrick says: Amen. What a great takedown of our data-obsessed culture, and reminder that there are still situations, like elections or anything personal in nature, where data doesn’t tell nearly the whole story.
Derrick says: This is one of the best arguments you’ll hear for why the cloud is an inevitable home for data. It’s so cheap to store and so easy to analyze huge amounts of data, that the economics and effort of on-prem systems just don’t make sense.
Derrick says: This interview with Uber engineer Susan Fowler is quite good. Companies just getting started with microservices, or running them at small scale, should take heed of the advice large users provide—because stuff can get out of control very quickly.
Derrick says: This seems like really good advice for companies as they rush headlong into AI. Slow down, put some process in place and maybe figure out data science first.
News from underneath your fave applications
Derrick says: This is a fascinating read if you want to learn the history of Twitter’s infrastructure, from the network to storage to its Puppet usage. There are probably some lessons to be learned in here too, ranging from engineering decisions to the business decisions about how much to invest in infrastructure and when. It also feels like an attempt to capture all of this in one place before it’s too late. 
Derrick says: Listen, we all have a love-hate relationship with Uber, but it looks like we’re going to have to start respecting as a company on the cutting edge of software engineering. Singhal is just the latest in a series of big hires for Uber’s infrastructure and engineering teams. And the company is trying to solve some unique problems.
Derrick says: What more is there to say? Another Facebook data center, another step up in efficiency—this one is powered entirely by renewables. How sustainable is Facebook, you ask: It also is now sourcing its own environmentally friendly plastic for data center gear. 
Legacy IT still has money ...
Derrick says: It’s a deal potentially worth $62 million over five years, but what’s the cost of missing out on the public cloud revolution? This type of deal with a conservative institution speaks to the challenges faced by legacy software vendors.
Derrick says: The shift to the cloud and to application containers is real, and the percentage of customer workloads not using some new architecture and running on some new type of platform will continue to shrink. Purchasing SimpliVity might buy HPE some time to figure out its cloud strategy, but it is not a recipe for relevance over the long term.  •  Share
Derrick says: This probably will not end well for Apiary’s customers, nor will it make Oracle’s cloud any more palatable to anybody but Oracle customers. Of course, there are more than a few of those to sell to.
Did you enjoy this issue?
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
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.