ARCHITECHT Daily: Artificial intelligence: Interviews, innovation and the economy

I came across so much good content on artificial intelligence between Friday and this morning, it's h
ARCHITECHT Daily: Artificial intelligence: Interviews, innovation and the economy
By ARCHITECHT • Issue #71
I came across so much good content on artificial intelligence between Friday and this morning, it’s hard to know where to start. It’s also getting hard to reconcile the arguments—often from the same people—that AI will simultaneously turn our economy of its ear, potentially grow intelligent enough to threaten humanity and, of course, be great and bring great innovations. 
Perhaps the key to understanding why we get all the signals we get about AI boils down one word: capitalism. If something will drive profits for corporations, then they’re going to invest in it at all costs. Or perhaps it’s that many of the people working on, investing in and/or thinking about AI are simultaneously realists and optimists. The prevailing logic being that we can’t slow down progress in AI, but with enough attention to the risks, we should be able to harness it mostly for good.
It seems that what you never hear from AI thought leaders or researchers, despite so many doom-and-gloom predictions, is that we should put the brakes on it. If only to give business leaders and policymakers a little breathing room to think about the future they want to see. 
For what it’s worth, I don’t think we can or should slow down AI research, if only because advances in fields like medicine, agriculture and climate science could turn out to be truly revolutionary. I also don’t expect (as regular readers have no doubt figured out) that AI is going to have extreme negative effects that lots of others predict. Some of those reasons have to do with technological limitations and human ingenuity, some have to do with legal/regulatory checks, and other have to do with a (perhaps very misplaced) optimism that people will eventually figure out that just because we can automate something, that doesn’t necessarily mean we should
If the populist wave sweeping across the United States and Europe really is about the economy, jobs and disillusionment, then, eventually, something has to give when it comes to automation. There are plenty of very bad outcomes for the future of Earth that won’t emerge from AI research labs, but rather from billions of people feeling hopeless.
Now that you have my two cents, check out this collection of news and interviews with much smarter folks on what they want, fear and expect from AI.
Warren Buffett has mixed emotions about AI
About the jobs and the global economy …

Sponsor: Cloudera
Sponsor: Cloudera
Artificial intelligence
I wrote about Amazon’s ambitious AI plans here, but in case you want another source, check out Jeff Bezos’s recent comments at a Seattle tech conference.
Good. If we could accurately predict the outcome of sporting events, elections or other events affected by extra-digital factors, they would eventually go away. As the saying goes, “That’s why we play the games.”  •  Share
As I was just saying … 70 percent accuracy is good, but there’s no doubt some important decisions in the 30 percent the model couldn’t predict. And times, justices and political climates change.
This time, researchers are using deep learning to analyze brain scans in order to figure out how brain activity correlates with seeing certain images.
  1. Drug and Materials Discovery
  2. Supply Chain & Logistics
  3. Financial Services
  4. Artificial Intelligence
  5. Cloud Security
These are good discussions to have, because most people (as with general AI) will want to think about the opportunities. The most amazing research doesn’t mean a whole lot if people don’t know enough to be excited.
The black-box nature of many AI models is a topic that I have written about and covered here on numerous occasions, and which other folks are writing a lot about, too. I consider essentially an extension of the correlation-versus-causation debate that has long-permeated discussions about big data (and probably all statistics before that). There are areas where we probably shouldn’t care what’s happening inside the model (e.g., product recommendations) and times when we should (e.g., anything involving life/death, civil rights or scientific research). Here are two good posts on that topic:
The state of explainable AI (Jillian Schwiep / Medium)
Listen the the ARCHITECHT Show podcast. New episodes every Thursday!
Listen the the ARCHITECHT Show podcast. New episodes every Thursday!
Cloud and infrastructure
Despite the happy talk about Kubernetes-OpenStack synergy, you have to wonder if Kubernetes advancements won’t eventually push OpenStack out of the picture and replace it with bare metal. With enough cross-project collaboration, OpenStack’s deep pockets could help that happen.  •  Share
There are so many reasons to think ARM in the data center might finally catch on, but equally as many to think it won’t. Red Hat could play an outsized role, if the right incentives, or demands, are in place.
If we’re still debating whether ARM has a meaningful future inside servers, what are the chances RISC-V catches fire? You could argue it’s getting pinched from x86 and ARM on one end, and AI/IoT chips on the other.
I read this paper last week and decided against posting it, because I didn’t find the Docker vs. bare metal comparison too compelling (of course Docker will add overhead, right?). But this Hacker News discussion about it is decent.
On its face, this is an argument against using Node.js to build serverless functions. Really, though, 
Media partner: GeekWire
Media partner: GeekWire
All things data
I don’t like this metaphor being trotted back out as if it’s new, but I do like this 2017-era examination of it. What we’ve seen happen is lots of silos, consolidation of power among major internet companies. What has to happen to keep “big data” from becoming the next “big oil?”
This is one of those stories about data-for-good that you really have to hope works out. Data sharing has proven more difficult than I think a lot of folks predicted, but maybe NYC can provide a workable and repeatable model. 
A good analysis of trends surrounding Apache Kafka, including from a survey just released by Confluent. I might take issue with the notion that cloud-provider solutions will dent Kafka adoption too much, though—at least if open source and containerization keep growing, too.  •  Share
This is a good approach to introducing machine learning to a platform, by focusing on a specific use case—in this case, time series anomalies. On an unrelated note, Elasticsearch gets most of the attention, but Kibana is a really powerful component of the Elastic / ELK stack.
This paper is a good argument for understanding the importance of data gathering and munging/wrangling, and perhaps for some standard levels that will help organizations gauge where they’re at.  •  Share
You could argue, as the authors of this paper do, that it does. You could even present some alternative frameworks, as they do. But what would it really take for the Spark community, for example, to embrace something new?  •  Share
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
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.