The 4 stories you should read today: From layoffs to recommendation systems

I'm keeping it light again for the last issue after a slow news week. Here are four (actually five) t
ARCHITECHT
The 4 stories you should read today: From layoffs to recommendation systems
By ARCHITECHT • Issue #108
I’m keeping it light again for the last issue after a slow news week. Here are four (actually five) things I think were really interesting or newsworthy, but about which I just don’t have a lot to say.
China’s Baidu being probed after CEO tests driverless car on public roads (Reuters): Apparently authorities are investigating whether Baidu CEO Robin Li conducted part of his keynote at Baidu’s big conference Wednesday sitting inside an unlicensed driverless car. Any actual rule-breaking with regard to permission is probably a very small deal, but suggestions that the car wasn’t following rules of the road might be noteworthy. I’m sure that’s easily fixed in production vehicles but, you know, the whole point of driverless cars is that they’re supposed to be better than humans.
Microsoft to layoff 3K sales employees, will focus on cloud (Fortune): That’s a lot of jobs, which is both sad and another result of the shift to cloud services. Theoretically, cloud computing should help streamline the sales process on both sides of the equation, which is not a good situation for salespeople. I’m not certain it’s a sign of anything wrong with Microsoft, but rather that Microsoft is gearing up for a tough international fight for cloud business. That will require some different sales skills faster cycles.
There’s a fight brewing between the NYPD and Silicon Valley’s Palantir (BuzzFeed): Palantir is involved in multiple legal situations at the moment and, frankly, things do not look that good from the outside. But Palantir’s business problems aside, there’s a bigger technological story here about the potential pitfalls of proprietary systems. The root of the showdown is that NYPD moved to a system comprised of IBM tools and its own home-built software, which it claims give the department better features and more control. Now NYPD wants Palantir to hand over its data stored on Palantir’s system, in a format the NYPD can use, and there’s disagreement over whether that happened.
Recommendations systems are an unsung hero of the web: Seriously, we interact with them so frequently, and they’ve driven God-only-knows how many sales on Amazon, views on Netflix and clicks on BuzzFeed. With that in mind, here’s a good breakdown of several popular methods (including deep learning) for building them (from Stats and Bots), and here’s a look at Amazon’s evolving and diverse approach through the years, from collaborative filtering to machine learning (from IEEE).

Sponsor: CircleCI
Sponsor: CircleCI
Listen to the latest ARCHITECHT Show podcast
In this episode of the ARCHITECHT Show, IBM Fellow and Watson Data Platform CTO Adam Kocoloski talks about the evolution of big data—from his days as an MIT physicist and co-founding Cloudant, to the application design trends dictating today’s cloud data platforms. In between, Kocoloski touches on a variety of topics, including the fate of Hadoop, the promise of quantum computing, the role of specialized hardware for AI and big data, and the effects of marketing on selling the Watson technologies. 
Sponsor: Linux Foundation
Sponsor: Linux Foundation
Artificial intelligence
It’s odd that some people consider this a “smart speaker” market, but I suppose the rule is to follow Apple’s guidance. However, even with Alibaba, Samsung and some third-party vendors promising products, it’s hard to see how anybody dethrones the vertically integrated platform players (Amazon, Google, Apple).
This an interesting article, touching on everything from coal plants to wind farms to data centers. Forget consumers: If there’s an area where AI can really change the world, it’s by helping stem climate change.
www.bbc.com  •  Share
Buried within the us vs. them narrative here, there are some good points about why AI, and big data in general, have struggled to master agricultural applications. However, AI will change farming eventually, which actually could be very good news for farmers.
I shared a link to this paper yesterday, but here’s a more consumable explanation by one of the Facebook engineers behind ELF. We’ve seen similar platforms from DeepMind and OpenAI, but of course their real value won’t come from training game models, but rather applying the techniques to new fields.
A group of Notre Dame researchers built a framework called SHADHO (Massively Scalable Hardware-Aware Distributed Hyperparameter Optimization) for matching tasks with the hardware in messier computing environments.
arxiv.org  •  Share
So, we should just shut down all online discussion, right? 
arxiv.org  •  Share
This is yet another step toward truly intelligent AI (along with research into context-aware systems, and systems with better memory, etc.). Read the third paragraph of Section 1 for a clear analogy on what the team here is trying to do.
arxiv.org  •  Share
Sponsor: Bonsai
Sponsor: Bonsai
Cloud and infrastructure
I hadn’t noticed this partnership, but it is kind of intriguing. The headline writer in me loves the notion that both are working against Google as a common enemy on autonomous cars, AI and cloud computing.
It’s AWS’s 62nd price cut, but who’s counting? Depending on size and region, the prices have dropped by more than 50 percent.
If you’re into that sort of thing—which more companies will be as they start trying to deliver products around AI and IoT, and even microservices, where speed matters and there’s a lot of of data traversing the network.
The company announced this a while ago, but this reminder is timely, because I’m sure lots of people forgot about it. It’s another reminder that the network matters a lot when you’re talking about distributed cloud services.
Sponsor: Cloudera
Sponsor: Cloudera
All things data
It’s not just for AI, either. National labs and supercomputer centers, as well as institutions like NASA and CERN, have so much data it’s hard to balance power, speed and cost across the board.
You might not have heard of StreamSets, but this interview with its CEO is a good introduction to its data integration technology and the thinking behind monetizing big data companies.
medium.com  •  Share
I did not, so I found this interesting at a high level. The blog post itself is about ensuring consistency when using Hadoop backed by S3.
I haven’t listened to this yet, but am excited to do so. As I’ve written several times here over the past few months, Pinterest seems like an optimal environment for doing data science and machine learning.
Sponsor: DigitalOcean
Sponsor: DigitalOcean
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.