ArchiTECHt Daily: Google's uncomfortable turn as enterprise IT vendor

Look, I understand that enterprise customers bring in the big bucks and that Google needs those compa
ARCHITECHT
ArchiTECHt Daily: Google's uncomfortable turn as enterprise IT vendor
By ARCHITECHT • Issue #33
Look, I understand that enterprise customers bring in the big bucks and that Google needs those companies on board if it expects its cloud business to be huge. However, I also understand that a company has to be what a company really is—what’s in its DNA—and Google has innovation encoded in its DNA.
Do you remember a few years ago, when Google co-founder Sergey Brin skydived onto the roof of the Moscone Center, live-streaming the fall via Google Glass? That’s Google. That’s the kind of vision I expect to hear from every Google business, even when the ultimate audience might be CIOs instead of consumers.
So it was a little awkward when Google Cloud SVP Diane Greene kicked off the opening keynote of the company’s Next conference on Wednesday by gushing over a new partnership with SAP (as if it doesn’t also have partnerships with AWS and Microsoft). And then trotting out enterprise customer after enterprise customer, none of which (save for maybe HSBC) had anything particularly interesting to say about what it’s doing or why it could only do this in Google’s cloud. Colgate-Palmolive employees are sharing files in Google Drive? Mind. Blown.
The keynote only really got interesting about halfway in, when Google Cloud chief scientist Fei Fei Li came out, speaking excitedly about artificial intelligence and announcing the company’s new Video Intelligence API, as well its acquisition of data-science competition platform Kaggle
And then when Alphabet chairman Eric Schmidt came out and actually shared Google’s broader vision for cloud, exhorting companies to “just get to the cloud now"—like in months instead of years. He gave the first mention of Google’s Cloud Spanner service that generated so much buzz last month, and the first meaningful mention of Kubernetes, and of Snap’s engineering accomplishments on Google App Engine. 
(Unless there’s some big surprise waiting in the wings, Cloud Spanner’s announcement generated more excitement than, I would predict, everything that Google will announce at Next on Thursday combined. Why not save that for your flagship cloud event to really get folks excited? In fact, a lot of really cool Google services—including ones that would be very good for enterprises, like Customer Reliability Engineering—get dribbled out via blog post and probably don’t get the attention they deserve.
On a related note, why was a Q&A with Vint Cerf and Marc Andreessen on the future of computing relegated to a lunchtime affair instead of being part of the keynote? )
I wrote last week that I believe Microsoft and Google can challenge AWS over the next few years, but that will require each company staying true to itself and giving customers an easy way to distinguish why they’re different and better than AWS. AWS is still the 800-pound gorilla in cloud computing and a very clear vision for enterprises, as well as for the army of startups—some of which became very successful running on AWS—that built AWS into the juggernaut it is today. Just go and watch the keynotes from its re:Invent conference in December.
You don’t win that fight by standing toe to toe and trading blows. If Google’s big advantage is its kick-ass infrastructure, engineers and computer scientists, then now is the time to hear about them. CIOs know they can run their basic workloads on Google Cloud, that’s not up for debate. Want to wow them? Show how customers are using Google’s expertise to drive efficiency and reliability to levels they couldn’t pull of anywhere else. 
And certainly use your time in the limelight to highlight all the innovative, next-generation stuff Google Cloud can power for companies that want to change the world, not just for those that want to save a few bucks on opex or sell a few more hammers. That’s the kind of vision I would have expected to see, because that’s Google’s DNA.
I’ve never really been one to judge a company on a conference keynote but, then again, I don’t recall ever seeing one that felt so incongruous to the company itself. If I were a developer or a startup entrepreneur—someone, say, trying to build the next Snap or, as Andreessen discussed in the Q&A, the next big breakthrough in genomics—I’m not sure Wednesday would have delivered the message that I’m part of Google Cloud’s target audience.
P.S. As I’m getting ready to hit "publish” on this, Google’s Urs Hölzle is on stage talking about Cloud Spanner and announcing new cloud-instance types and billing models. Not exactly visionary stuff, but the customer case study from Schlumberger about doing HPC with GPUs is pretty cool (even if it’s for oil exploration).

Sponsor: Datos IO
Holy hardware revolution from the Open Compute Project!
If Microsoft begins rolling this out across its fleet of Azure servers, that would not be good news for Intel or, let’s assume, server vendors like HPE. Longer-term, Linux on ARM servers could be an even bigger deal.
Microsoft wasn’t the only newsmaker at the Open Compute Summit yesterday. The link above highlights everything that Facebook announced. And here’s the rest in more detail:
One is the cloud-provider-focused HGX-1 Hyperscale Accelerator linked to above, and the other is the new Jetson TX2 that’s designed for running inside drones and other embedded endpoints. 
Good news for white-box partners like Wiwynn, not so good news for partners like HPE that would love to not sell commodity gear.
Around the web: Artificial intelligence
Wow. Uber only acquired his startup, Geometric Intelligence, in December. I can’t imagine that was the plan.
I spoke about this yesterday after the rumor surfaced. Here’s the proof.
As well as updates to some existing AI products. But the vision thing is really quite cool for identifying objects in every single frame across a video, or a library of videos. 
It claims a new system hit an all-time best 5.1 percent error rate. Here’s the paper: https://arxiv.org/abs/1703.02136
I hope this guy keeps writing these synopses of recent, and big, AI papers, because it’s really good.
medium.com  •  Share
I like this quote from Accel’s Jake Flomenberg: “Most AI startups are not going to fail on the basis of their actual AI. They fail because they fail to identify a problem that needs solving.”
Depending on how broadly you define AI, I would have guessed there’s be about 100.
The link above is about more efficient object recognition in video streams using some database principles. This one is about “active learning” across smaller datasets. This one is about making models work better in the real world using adversarial learnings.
arxiv.org  •  Share
This is actually not AI at all. It seems more like standard machine learning, and it’s kind of refreshing to see that someone doesn’t feel compelled to use AI in the name being hip. 
Sponsor: Onehub
Around the web: All things data
Like Cassandra, but targeting developers and, the company claims, much faster.
One bit on one atom. It’s a pretty remarkable achievement, but nowhere near ready for life outside the lab. And, remarkably, the ultimate use of this technology might not be data storage.
Yeah, Teradata is doing open source now. That’s probably a good thing, but it’s probably not particularly good for Teradata over the long run.
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.