ARCHITECHT Daily: Whole Foods as a treasure trove of data for Amazon

The announcement on Friday that Amazon is buying Whole Foods set off a wave of takes about Amazon vs.
ARCHITECHT Daily: Whole Foods as a treasure trove of data for Amazon
By ARCHITECHT • Issue #98
The announcement on Friday that Amazon is buying Whole Foods set off a wave of takes about Amazon vs. Walmart and Amazon’s 400 new distributions centers for grocery delivery, but Tom Krazit at GeekWire nailed another major aspect of this deal: Amazon now has a huge data source on brick-and-mortar shopping behaviors that were previously a blind spot to the company. And, as I wrote last month about Facebook, data reigns king in the age of artificial intelligence.
Amazon used to know a lot about how people shop online and how its e-commerce distribution channels operate. With Alexa, it’s also learning a lot about how people talk, what they put on shopping lists and, more generally, what they’re interested in. Using all that data, it’s already using machine learning to optimize and improve these businesses.
Now Amazon will also a lot about how people shop for groceries in the physical world. It will know what they buy and when, what they buy together, and what they buy in-person that they don’t buy online (even if they could). It will understand how grocery stores operate and how grocery distribution channels works. And then it will use machine learning to optimize these things, too.
I don’t know if Seattle-area neighbor Costco is quaking in its boots, but other grocery chains probably should be. Should Amazon push Whole Foods to revamp in terms of pricing and product selection, they are about to, to use a politically incorrect term, get Ubered. Amazon had the AI expertise, and now it will have the data to make this happen.
Instacart, whose customers shop at Whole Foods a lot, has been publishing some interesting blog posts lately about shopping data, deep learning and grocery-delivery logistics. Given its tech-centric nature, maybe the correct analogy here is that Instacart is Snapchat, and that it could very well be Facebooked (or Instagrammed).
In his GeekWire post, Krazit also suggests that Amazon will now have the data and expertise to offer a cloud-based retail service if it’s so inclined. That’s certainly possible —the smart money is in vertical AI solutions, as Bradford Cross explained on the ARCHITECHT Show podcast last month, and recently in a conference talk—but Amazon would have to believe it can deliver this without cannibalizing its own business in the process. Amazon Web Services has been venturing much more deeply into the application space over the past couple years (more on that further down), but it hasn’t offered any vertical-specific offerings I can think of, much less in a space where Amazon already operates.
However, speaking of the cloud, another interesting data angle in the Whole Foods acquisition is that Whole Foods is a big Microsoft Azure customer. According to a story in The Register, Whole Foods is a major user of the Azure Active Directory service, which it uses to provide its 90,000 employees with single sign-on. It seems safe to assume an Amazon-owned Whole Foods will not stay on Microsoft’s cloud for any longer than it has to (although, FWIW, Zappos was still struggling with a migration to AWS as of late 2015) but any data Whole Foods has on where the Azure service excels and lags could also be valuable.
Like many people, I have mixed emotions about Amazon in terms of what it wants to do and how it goes about doing it. But I’m thinking about it a lot differently today than I was on Thursday. To me, it’s now the ultimate guinea pig for the power of AI to transform business—a cliche we hear a lot but that has arguably never been so publicly put to the test. The economics of the grocery business have been well-understood for decades, so we’re going to see firsthand how Amazon’s approach changes things.

Sponsor: Cloudera
Sponsor: Cloudera
Listen to the latest ARCHITECHT Show podcast
In last week’s episode, I interviewed Cloud Native Computing Foundation executive director Dan Kohn, who discussed the Kubernetes ecosystem and adoption, as well the purpose for foundations like CNCF and where they fit into the open source landscape (e.g., in relation to the Apache Software Foundation).  I also spoke with Gabe Monroy, who was co-founder and CTO of container startup Deis, which Microsoft acquired in April. Monroy talked about how Deis came to be and the technologies it developed, and how everything is working now that they’re part of Microsoft.
Artificial intelligence
I linked to this post from Bradford Cross above, but here it is explicitly. I’m not entirely sold on this argument—I think there will be big business for companies that can help enterprises build bespoke solutions—but it has a lot of merit.
To me, this whole discussion seems to be more about Apple’s business than its AI capabilities. If Apple keeps moving units, people will use CoreML to build AI-powered apps for those devices. But Apple will never rival Google or AWS in terms of innovation or broader developer reach.
This has some good info and analysis, including on current investment levels. I suspect we’ll see computer vision drop in terms of overall percentage soon enough. Also, Canada is not listed as an AI hub, which is wrong.
This is a good podcast interview with Tim Hwang of Google, especially the parts about interrogating your data for signs it might be biased, and about acknowledging that technologists are rarely policy experts.
Yes, Ant Financial is doing some interesting things and Alipay is hugely popular in China. But let’s not read too much into what its success means in comparison to Apple Pay, for example. Different tech for different cultures.
This interview covers a lot of ground, from talent demands to blackbox algorithms. Here’s a taste on the former: “I was recently trying to find someone to come and consult for a big company … I just could not find anybody who didn’t have some [competing corporate] involvement.“
Roboticist Rodney Brooks doesn’t think driverless cars are as imminent as many people seem to believe. These examples highlight both infrastructural and decision-making challenges that are worth chewing on if you’re working in this space.
This article provides a good roundup of research findings, as well as startup activity, that suggest AI will be a boon for health care.  •  Share
Seriously, I will pay money for this today. Security cameras that give more nuanced alerts would be super-helpful, and hopefully are coming as object- and action-recognition picks up.
A general-purpose AI model would be a big deal if it can be applied outside a research setting. Deep learning has helped reduce certain time-consuming elements of deep learning, but it’s still time-consuming.  •  Share
Also, it built a model that outperformed human-centric techniques at deciding which parts of a workloads should run on what machines. I’m sure this will find its way to a cloud near you at some point.  •  Share
Sponsor: DigitalOcean
Sponsor: DigitalOcean
Cloud and infrastructure
It’s not Google or Facebook, but Dropbox has had to grow its infrastructure to keep up with traffic, as well. Here’s how it went from cloud to its own data centers to its own edge network.
We now have more insight into how AWS decides to pursue non-compete actions (it really is on a case-by-case basis) and some evidence that former Google VP Adam Bosworth is heading the AWS project at issue here.
It’s a pilot partnership around a service called CarData, for storing vehicle data and making it accessible to third parties (e.g., insurance companies or mechanics). At any rate, it’s a win for IBM.
IBM OpenWhisk and Fission are making waves, but startup Serverless and its eponymous framework for AWS Lambda are really killing it. Watching this space grow alongside the still-immature-in-its-own-right container space will be interesting.  •  Share
That leaves just Microsoft and AWS on the outside, right? Microsoft is certainly betting big on Kubernetes, but AWS is still pushing ECS as its container orchestrator of choice.  •  Share
China is the top two spots, and Switzerland is in third. It’s not a good look for the United States HPC community, but we still have a lot of powerful supercomputers. When will people stop caring so much about this in a world with cloud computing, quantum computing and scalable AI algorithms?
Sponsor: CircleCI
Sponsor: CircleCI
All things data
This is a not-so-great review of the Amazon Elasticsearch service, from someone who knows ES well. This is evidence of what many open source companies argue about cloud competition: They can’t/won’t build it as well as we can.
As far as I’m concerned, IoT is essentially a data play, so companies with an interest in capturing that data might take notice that Bosch is investing so heavily in driving (no pun intended) adoption of smart devices.
This is a noble cause from SAP Ariba, but the challenge, as with everything, is getting the right data and getting businesses to pay attention.
Sponsor: Bonsai
Sponsor: Bonsai
If you read FlowingData, this is founder Nathan Yau’s recent commencement speech to the UCLA Statistics school. It provides good context on how important this field has come, and how far it still has to go.
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.