Dremio has a new take on analytics, open containers advance, and IBM is pushing its cloud biz

Let me start with an apology for the late newsletters this week. It's remarkable how changes in your
Dremio has a new take on analytics, open containers advance, and IBM is pushing its cloud biz
By ARCHITECHT • Issue #115
Let me start with an apology for the late newsletters this week. It’s remarkable how changes in your kid’s schedule can mess with yours, too. Tomorrow’s will likely be even later (or, more likely, delivered on Friday morning) because I’m flying up for Cloudera’s Wrangle conference in the morning. If you’ll be there and want to catch up, you can find me hanging around and taking in what looks like a really good agenda. (And I’m not just saying that because Cloudera’s a sponsor.)
I want to start today’s newsletter by pointing out that IBM continues to take heat—in this case, for more than 5 years of declining revenues—and then masterfully deflect attention onto the positives. The positives in this case being growing cloud revenues and four new cloud data centers. It also continues to defend criticisms of its Watson and AI businesses, which have disappointed many investors and people with strong opinions on AI.
What’s crazy is that if you look at IBM, you see lots of really interesting research and early investments in technologies that have the potential to be game-changers—including, recently, Kubernetes and blockchain. But something happens between there and sales that doesn’t add up. That being said, a growing cloud business is a bright spot (even if it includes everything but the kitchen sink). And, really, who’s actually making real money selling AI right now?
Anyhow, I’ve written about IBM recently and spoke with IBM Fellow and Watson Data Platform CTO Adam Kocoloski on the podcast. I think they’re both worth reading, especially the podcast highlights.
The other biggish infrastructure news on Wednesday is that the Open Container Initiative (which Tom Krazit at GeekWire rightfully points out “sounds like an outdoor keg party”) released the 1.0 version of its specification. I don’t have a lot to say about this, except to point out that the original goal of OCI was really to ensure that Docker didn’t end up owning the container market from top to bottom. Docker, to its credit, has played along very nicely because—IMHO—it’s still really the only container format that people use, and because embracing open standards (with OCI and the CNCF) is better for business than being on an island.
Here a bunch of blog posts about OCI 1.0, including two Docker-centric ones from Docker:
Finally, I want to point out that Dremio officially launched on Wednesday, after spending a couple of years in stealth mode. Dremio has built data analytics platform, based on Apache Arrow, that’s designed to more or less do away the enterprise data warehouse. I’m oversimplifying, but its technology aggregates data representations from across data stores (including HDFS, S3, MongoDB, Elasticsearch, RDBMS), stores them in-memory, and then connects to a slew of BI and data science tools.
I spoke with Dremio co-founder and CEO Tomer Shiran about the company, and couldn’t help but think his team (many executives came from MapR and MongoDB) learned a lot of lessons from the era of big data 1.0. Some of those lessons were on the UX front with decisions like investing heavily on UI, self-service and collaboration from the outset, but the biggest one might be realizing that there is no one data lake right now. 
“I don’t think you can build a multi-billion-dollar market doing [X for Hadoop,” Shiran said. 
Another big thing is owning the competitive nature of Dremio against traditional data warehouse and ETL approaches and companies. Whereas Hadoop companies began with a very partner-centric approach on those two industries, really only positioning the technology (to use Shiran’s term) as “a cheap data warehouse,” Dremio is clearly trying to cut out the middlemen. It’s a tall order, but the timing is right and it’s hard to disrupt entrenched markets by playing nice.

Sponsor: Bonsai
Sponsor: Bonsai
Artificial intelligence
Of course Apple, which is regularly dinged for not being open enough, has its own AI journal. But that aside, the first post is really quite interesting, and on a topic—creating synthetic images—that’s going to be super-important as companies and researchers try to train models using less real-world data. It can be hard to come by.
Nauto seems to be taking a smart approach to building driverless cars, by first retrofitting existing cars with technology to analyze driver behavior and improve safety.
I recall speaking with these guys a few years ago, and it appears the company has found a market for its operating system that can give some intelligence to dumb machines. Importantly, Qualcomm is an investor and Brain Corp. wants its software to run on Qualcomm chips—first Snapdragon and then on AI-optimized chips.
Part of me is burned out on “AI” for virtual assistants, but something that actually works and fits into the existing workflow of managing emails, meetings, etc., would be a godsend for lots of people. Clara is starting out by using humans to work on uncertain aspects.
After reports recently that Samsung was struggling to master Bixby’s English speech recognition and NLP, the feature is finally making its way onto U.S. devices. Other English-speaking countries are still on the waiting list.
I like that rather than grilling companies or fear-mongering, U.K. lawmakers appear to be asking open-ended questions and looking for opportunities to exploit as well as potential problems to think about.
Speaking of regulations, AI researchers are not happy with Elon’s Musk telling the National Governors Association we should be afraid of AI. The headline links to comments by Rodney Brooks, who was the founding director of CSAIL at MIT, and here’s a collection of other experts, including Fei-Fei Li and Oren Etzioni.
It’s authored by Erik Brynjolfsson and Andrew McAfee of MIT, who are out promoting their new book. If you’re looking for a good, non-technical explanation of where we’re at and what’s possible with AI, this is the piece to read.
hbr.org  •  Share
Basically, goes the argument here, most companies aren’t yet ready to adopt AI at the same level as Google or Facebook, and they won’t be until they have at least already mastered big data. Also—and this is important—most companies won’t steal talent from Google or Facebook, so they’ll have to find it elsewhere, nurture it internally and expect help from vendors.
redmonk.com  •  Share
This is hardly the first approach to running neural networks on phones and other devices, rather than on GPUs in the cloud, but the researchers here claim their method cuts power consumption by 73 percent over traditional approaches.
Their system, called Houdini, is interesting because it attacks overall performance on tasks that aren’t simply classification. Also interesting is that they demonstrate it on Baidu’s DeepBench and Google Voice.
arxiv.org  •  Share
Sponsor: Cloudera
Sponsor: Cloudera
Cloud and infrastructure
Essentially, Oracle owns the machines and customers pay Oracle on a subscription model to manage the services running on them. For more on the services available, which range from big data to SaaS offerings, check out this post.
I’m still a sucker for a well-written vendor case study, and this one fits the bill. There’s a good section on the value of ecosystems, too, because all the customer’s vendors used AWS and there is lots of AWS talent out there.
This is a good how-to on securing Amazon S3 buckets, starting with the simple advice of not turning off the “private” default and making them public. Sadly, there have been multiple breaches lately (and not just on AWS) because companies have left data stores open.
This is a good read on the ebbs and flows of cryptocurrency on GPU sales, but it appears more strategic to short-term investors than to either company. Some think ASICs will ultimately win out for cryptocurrency mining, and gamers and data centers are more predictable markets.
Beyond the discussion of quantum computing technology here is a good discussion about the long-term value of data. There’s a fear that hackers are stealing encrypted data and banking on quantum computing eventually allowing them to decrypt it. But some question how much value years, or even decades-old, data really has.
Sponsor: DigitalOcean
Sponsor: DigitalOcean
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?
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.