AMD wants to challenge Intel, and a quantum startup wants to challenge Google

I'm going to keep this brief, as it's late and I have an early morning flight. So here are the most i
ARCHITECHT
AMD wants to challenge Intel, and a quantum startup wants to challenge Google
By ARCHITECHT • Issue #99
I’m going to keep this brief, as it’s late and I have an early morning flight. So here are the most interesting stories of the day, as far as I’m concerned (and, yes, there’s kind of an underdog theme happening):
AMD wants to challenge Intel in the data center: AMD unveiled its new Epyc data center chip today, to quite a bit of fanfare. Folks seemed to generally have good things to say about the Epyc design but, of course, the market for server chips has changed drastically since AMD was last a real player (read: “the cloud happened”). In that regard, it’s probably meaningful that Microsoft Azure and Baidu have pledged to offer Epyc-based instances at some point, although—assuming customers actually want to adopt a new CPU architecture—how cloud providers price them and how widely they make them available will dictate how successful those partnerships are. Also, it would help if AWS offered AMD instances, and if the trend wasn’t toward managed services and container- and function-level abstractions.
Rigetti Computing wants to challenge Google and IBM in quantum computing: If you haven’t heard of Rigetti Computing, now would be a good time to get familiar with it. The company, which has raised more than $64 million (led in two rounds by Andreessen Horowitz), just announced version 1.0 of its Forest platform for writing and executing quantum algorithms, as well as a fabrication facility for manufacturing quantum chips. WIRED has a good writeup on the company around this news, too, including on the daunting challenge of not only developing a commercially viable quantum computer (this is different than D-Wave’s system; for more on the difference, check out the podcast link below) but also to beat IBM and Google to doing so.
MongoDB is stepping up its product game: Of all the well-known open source companies to have emerged in the late ‘00s, perhaps no one has had to endure as much criticism and speculation (about its financial situation) as MongoDB, whose eponymous database was both widely adopted and seemingly difficult to monetize. The company sought to change that with its hosted Atlas database as a service last year, and on Tuesday announced two potentially useful (and profitable) new services. One is a product called Stitch, for integrating MongoDB directly to popular data sources (e.g., Google, Facebook, AWS, Twilio, Slack and anything with a REST API), and an as-yet-unavailable analytics tool called Mongo Charts. 

Sponsor: DigitalOcean
Sponsor: DigitalOcean
Listen to the new ARCHITECHT AI podcast
In the inaugural episode of The AI and Robot Show, Derrick Harris speaks with Kindred.ai co-founder and CEO Geordie Rose about a wide range of issues in artificial intelligence and quantum computing (Rose also co-founded D-Wave Systems in 1999). Among the topics Rose covers are how Kindred.ai is seeking to build artificial general intelligence and the importance of a physical embodiment in that quest; the commercial viability and challenges of different approaches to quantum computing; the ethics of human-AI interaction; and whether his Canadian homeland can really become the world’s AI epicenter. 
Artificial intelligence
This is kind of a reference architecture for building cutting-edge models, complete with datasets and architectures, but importantly is also modular so users can personalize models to their own data, etc.
This is potentially very useful, as lots of people really do think better out loud than on paper. I’d like to put it to the test on podcast transcription. In somewhat related news, Microsoft also is backing an AI accelerator in Paris.
This isn’t exactly a new language, as some have suggested, but more like gibberish that produced a good result. That being said, we don’t really know what was happening underneath that broken English.
Can we be done with “smart” toys yet, or with emotion recognition? Save for extreme situations, I don’t think we need toys or anything else monitoring how we’re feeling at any given moment (as if we didn’t know ourselves).
If you’re looking for a big-money application of machine learning, it turns out the government is looking for a way to automate the process of awarding contracts.
gcn.com  •  Share
This is interesting technology, especially if it can actually discern meaningful patterns from unstructured and alternative data sources. Seems like there are important applications beyond financial markets, too.
If I read this correctly, the author is proposing that America should probably spend less energy trying to restrict Chinese investment in American companies and more energy (and money) making sure America is operating at peak innovation.
Sponsor: CircleCI
Sponsor: CircleCI
Cloud and infrastructure
The headline links to a story about how AWS is growing its presence in Canada, in terms of headcount and data centers. And here, Amazon CTO Werner Vogels blogs about the new AWS region in Hong Kong.
This interview with Cockroach Labs CEO Spencer Kimball provides some good insights into how the company modeled CockroachDB on Google Spanner, and how it uses an “honor system” for enterprise licensing.
medium.com  •  Share
… Chinese startup PingCAP just raised nearly $15 million for its take on a globally distributed SQL database. It’s based on an open source project called TiDB.
The only thing is, it’s not telling what the discount is upfront, for these bare metal instances running older-model Xeon processors. As the story points out, it’s a strange tactic and the reasons for doing it this way aren’t clear. Spot pricing this is not.
BeyondCorp is a cool approach to network security and identity and access management. FWIW, there’s a also a startup called ScaleFT trying to productize this approach.
We’re talking about aesthetics (e.g., rooftop pools), not just how servers are cooled and managed. As edge computing takes off, it does make some sense to think about design and mixed-use options for urban data centers.
Of course Cisco is offering a product that use machine learning to analyze traffic, but the subscription billing on its new products is probably the real game-changer in terms of Cisco’s business.
Sponsor: Bonsai
Sponsor: Bonsai
All things data
Julia markets itself as, essentially, a data scientist’s dream come true. It’s open source, fast, and has an MIT pedigree and lots of big-name users.
Like all interviews with Kafka co-creator and Confluent co-founder Jay Kreps, this one is very insightful and you should read it if you’re interested in data infrastructure and stream processing. (You also should read this.)
This blog post is a good reminder of what goes into building a SaaS service that’s different from managing something yourself. In this case, it’s a data pipeline custom-built to do a specific task(s) at massive scale.
German researchers developed a system that builds 4-D models of cities (including Las Vegas!), time being the fourth dimension. They say they can track changes, say in a building sinking, as small as 1 millimeter per year.
www.tum.de  •  Share
This is cool research out of Harvard into building a data system that can adapt different type of data over time (e.g., key-value and columnar data). The idea being that we shouldn’t have to spend so much effort building and rebuilding data architectures as data types and applications change.
arxiv.org  •  Share
Sponsor: Cloudera
Sponsor: Cloudera
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.