3 things to read today: From Rackspace's MSP acquisition to human-robot interaction at Amazon

On a relatively slow news day (if your beat doesn't involve covering iPhones), here are three stories
ARCHITECHT
3 things to read today: From Rackspace's MSP acquisition to human-robot interaction at Amazon
By ARCHITECHT • Issue #136
On a relatively slow news day (if your beat doesn’t involve covering iPhones), here are three stories you don’t want to miss:
Rackspace announces agreement to acquire Datapipe (Rackspace): Hey, if you’re going to quit the cloud business to focus on managed services, you might as well become one of the biggest MSPs around. Rackspace is definitely the bigger party here in terms of revenue, but Datapipe does bring some pretty strategic assets to the table. Rackspace and Datapipe both have pretty strong offerings lined up around the Big 3 public clouds, although Datapipe offers managed services for Alibaba Cloud, as well as traditional colocation services and 29 additional data center locations, including in mainland China, Russia and Brazil. Managed services aside, I also think Rackspace has its eyes set on owning the private part of hybrid cloud deployments.
As Amazon pushes forward with robots, workers find new roles (New York Times): I don’t know if it’s just wishful thinking, but this story follows in a line of research and analysis lately suggesting that there’s little evidence robots will actually take too many human jobs. In this case, as Amazon has increased the number of robots in its warehouses, humans have transitioned to new jobs (and arguably better ones), but haven’t lost jobs. It’s ironic to here this coming out of Amazon, because people also used to believe its cloud computing division would kill all sysadmin and ops jobs, but that clearly did not happen. What not entirely clear, though, is how these robots affect the hordes of Amazon seasonal contract workers we used to hear so much about.
How neural networks think (MIT News): This is more cool research into figuring out why models make the decisions they make—in this case by analyzing the effects of different inputs on a model’s outputs. One particularly interesting insight here is about just how important it is to pair large datasets with large-enough models. Run an underpowered model, and you end up with a dialogue system that automatically answers “I don’t know” whenever it sees a question beginning with “Who” or “What.”

Sponsor: Bonsai
Sponsor: Bonsai
Artificial intelligence
Technically, they’re learning to grab new types of objects without being expressly programmed or trained on how to do so. This isn’t a huge deal on its own—and it’s all still in the lab—but the domain of transfer learning and other attempts to generalize AI models could be a game-changer.
This isn’t a wholly crazy hope, although I think it depends on what they’re hoping to do. Computer vision on satellite imagery? Sure. Finding meaningful patterns in mountains of digital surveillance data? That seems a little trickier.
phys.org  •  Share
Because, as we’ve talked about here numerous times, the resources required to gather and clean data might be significant on their own. And that’s even before building and powering the models.
It’s not just a matter of copying, but actually a matter of learning how the levels are set up, how bad guys appear, etc. 
This article has some interesting stuff about Russia’s relative strengths and weaknesses in AI research compared with the U.S. and China. But I still think there’s a disconnect between what’s possible and what people imagine is possible. If anything, the real value of AI in this space might be its ability to do simple things at a massive scale.
This gets at the same thing as the MIT research linked to up top, only it’s focused on regulated industries such as financial services. The extent to which the “black box” problem limits adoption is shrinking pretty fast.
arxiv.org  •  Share
So mark your calendars for Sept. 13.
Sponsor: DigitalOcean
Sponsor: DigitalOcean
Cloud and infrastructure
Rackspace wasn’t the only company acquiring a data center footprint on Monday. It looks like Equinix is trying to capture the next phase of cloud adoption and build-out on the edge and in developing economies.
AWS and Google already offer similar instance types, so this is a bit of catch-up on Microsoft’s part.
This blog post has lots of details on how and why you’d want to use it versus other options. But the bigger picture is how the network is becoming a point of competition among cloud providers.
I don’t entirely agree with this take, which is that VMware is so important it doesn’t need to focus too heavily on a container strategy. The cloud didn’t kill VMs because it’s comprised of VMs, but containers, functions, and everything else coming down the pike are different beasts.
This analysis puts container and Kubernetes usage both at over 50 percent, and Cloud Foundry right around 50 percent. Some companies are using all three. Would be interesting to see where Mesos plays here, too. Also, this kind of contrasts with findings from a Cloud Foundry survey, which claims that production usage of containers in large enterprises is still well under 50 percent.
redmonk.com  •  Share
It’s basically some of the software startups and systems integrators you’d assume, as well as some less-household-names.
cncf.io  •  Share
The headline pretty much sums this up. It’s a good post on how we got from servers to serverless, and how each step along the way is different.
And stock price and market cap. Of those, cloud revenues will be the easiest to fudge.
This is a bit weird because Microsoft also has an IoT management platform last I checked, but I’m sure it will happily take whatever revenue it can from AT&T customers consuming Cosmos DB and Power BI resources.
That was my takeaway from this look at how the teams building the Summit supercomputer need to account for its huge power draw and heat output. You have to wonder when the world will decide that enough is enough when it comes to computing power, or when we’ll have a technological breakthrough that really lets us reduce energy consumption.
A former Facebook engineer explains the company’s thinking when it tied a “do not sue us for patent infringement” clause to the license of its popular React software. The desire to avoid burdensome patent litigation is certainly understandable, but this type of clause would be unworkable if it were in place more broadly among OSS projects.
medium.com  •  Share
If you enjoy the newsletter, please help spread the word via Twitter, or however else you see fit.
If you’re interested in sponsoring the newsletter and/or the ARCHITECHT Show podcast, please drop me a line.
Use Feedly? Get ARCHITECHT via RSS here:
All things data
This is a long (perhaps too long) story on Teralytics, but does raise some good points about the value of data on where people are going, as well as on the delicate balance of tackling real problems (e.g., infrastructure planning) versus doing location-based ads.
But the real thing I wanted to point out is that there’s a company commercializing Apache Flink—an alternative to Spark for real-time processing—and it is Data Artisans.
I’m not sure there’s any immediate commercial applicability here, but this research hits on the broader trend of GPUs being applied a much broader array of uses.
arxiv.org  •  Share
Naturally, building something specifically for a certain application should tend to provide better performance and ease of use. And something tells me there will be a lot more research done on blockchain before it’s as ubiquitous as some supporters believe it will be.
arxiv.org  •  Share
New ARCHITECHT Show every Thursday; new AI & Robot Show every Friday!
New ARCHITECHT Show every Thursday; new AI & Robot Show every Friday!
Did you enjoy this issue?
ARCHITECHT
ARCHITECHT delivers the most interesting news and information about the business impacts of cloud computing, artificial intelligence, and other trends reshaping enterprise IT. 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.