It's "the end of the beginning" for AI

There were two conferences in San Francisco last week that featured some powerhouse speakers from the
It's "the end of the beginning" for AI
By ARCHITECHT • Issue #102
There were two conferences in San Francisco last week that featured some powerhouse speakers from the AI world, including leaders from Google, Microsoft, Salesforce, OpenAI, Uber and UC-Berkeley. VentureBeat was on the scene at both and reported on a few seemingly interesting talks, all of which I’ll link to below. The connecting theme, however, appears to be, as Google’s Fei-Fei Li put it, that we’re at “the end of the beginning” with AI.
What that means, essentially, is that there have been some major breakthroughs and there has been enough value created that investments in AI will continue for the foreseeable future. There won’t be another “AI winter,” where funding vanishes because the technology can’t deliver on promises to customers, CEOs or university president. If anything, it’s actually difficult to keep up with the pace of advances.
On the other hand, there’s presumably a big gap between the beginning and any sort of end point in AI. As the researchers all noted, AI is not yet “intelligent” in the same way that humans are intelligent, and it will likely be a long while before that happens. In the meantime, we still need to solve with very real practical problems, such as building systems with at least a working method of contextualizing environments, and establishing regulations around, say, how certain is certain enough for applications such as driverless cars.
These are the types of messages I wish most people would get about AI, rather than the fear-stoking pieces about job losses and existential threats, or the breathless puff pieces extolling how AI will completely transform business as we know and turn the citizenry into one big leisure class. Aside from being great clickbait, these actually are all important issues, but there also are lots of smart people already thinking about how to address them. 
For everyday citizens and businesspeople, I think the bigger threat is actually believing the hype without having a real understanding of today’s reality (and what the term even means in most settings). It’s hard to see how we’ll be able to ride the AI wave smoothly—to take advantage of it meaningfully today and plan for a future where models might be much more powerful—when we’re drowning in a sea of hyperbole. At least the people actually working in the field seem grounded in reality.
Here are the writeups of those three conference sessions, which are short but good:

Sponsor: Cloudera
Sponsor: Cloudera
Listen to the latest ARCHITECHT AI podcast
In this episode of the AI and Robot Show, Signe Brewster speaks with Fetch Robotics CEO Melonee Wise about the state of the art in automation. Wise, who was also employee No. 2 at robotics pioneer Willow Garage, shares some details on how Fetch’s warehouse robots functions and also goes into depth on, among other things, human-robot interaction; how consumer demands and practices have influenced warehouse processes; the immense value of open source robotics libraries; and why the future of home automation might look more like the Roomba than like Rosie. 
Artificial intelligence
Algorithmia has been around for a few years and is an interesting, if not clearly lucrative, concept: a marketplace for machine learning and AI models and algorithms. Also, Google’s new AI venture arm led this round.
This is one of those fear-stoking stories I referenced above. It’s good to think about how AI will affect jobs and inequality, but let’s consider the chance that automation will be good for us, or at least that the apocalypse wouldn’t be preferable.
Among them: deep learning for computer vision, precision farming, and quantum computing.
A how-to from someone who works on the show about, explaining how they really built the app using TensorFlow and Keras. On a side note, the author of this post, Tim Anglade, also did a good series of interviews called “The NoSQL Tapes” several years ago.  •  Share
This makes a lot of sense, considering the embedded nature of so many military systems. But they’re also working with a $1.3 million grant to scale up neuromorphic clusters.
When Geoff Hinton gets written up in a weekend New York Times profile, you know AI has hit the mainstream. Speaking of Toronto, where Hinton resides, McKinsey & Co. has some advice for the Canadian government to carry out its AI ambitions.
The Objectifier is fairly useless, but also kind of genius. It’s a device that you train on an object or gesture, and it performs a simple task when it sees it. It’s a good way to get people thinking about how these models work. 
2 tons of Legos, to be exact. Normally I don’t share these hobbyist projects, but this one is particularly awesome.
A researcher working on a system called NFrame, with a $450,000 NSF grant, thinks they can. It seems more or less like anomaly detection, but perhaps faster and more advanced.  •  Share
Speaking of using to predict breaches, that’s SparkCognition’s game, too. I’m always intrigued when strategic investors are involved, because I assume they’re already seeing some benefit from it.
This is potentially interesting if you’re a researcher that wasn’t using Universe for some of the reasons highlighted here. However, the seeming lack of updates and progress on Universe highlight the larger issue of relying on open source tools without robust communities or financial incentives.
I linked to the original paper and blog post recently, about Google’s model that can do speech, text and object recognition. Here’s an article that talks about it at a higher level.  •  Share
Sponsor: DigitalOcean
Sponsor: DigitalOcean
Cloud and infrastructure
Yes, it’s possible Cisco can right the ship, but it will be remarkably difficult. The chart in this story is worth a thousand words.
At this point, companies like Google have most of the data (and the brains) about online activity, so who’s better equipped to secure it? I will take a benevolent dictator, or vigilante force, or whatever you want to call it, over anything slower, dumber or motivated by less-desirable goals.  •  Share
The Justice Department wants the Supreme Court to weigh in on the legal battle it has been fighting against Microsoft for the last few years. We desperately need updated—or at least clear— data privacy laws, especially in the cloud era, where data and users can be anywhere even if the company that can access that data is based in the United States.
Code management is becoming very bug business, and JFrog is one of the companies leading the charge. Cloudmunch monitors the pieces of this pipeline to identify problems.
The United States might have slipped down the Top500 supercomputer list, but being the first to an exascale system would make up for that. AMD, Cray, HPE, IBM, Intel and Nvidia are the grantees.
Sponsor: CircleCI
Sponsor: CircleCI
All things data
This project is part of a partnership with Cornell University, and part of a larger consortium around food safety. I swear, IBM Research works on a lot of projects that really make you root for IBM.
The original headline kind of speaks for itself. I think a lot of larger companies are struggling with the “single source of truth” piece, as they still are fighting with data silos. You’d like to think companies borne of the big data era will have an easier time dealing with this as they grow.
This is another 5-part list, only it frames the issue as learning big data best practices as a stepping stone to doing AI. This was the topic of my newsletter blurb on Thursday, as well.
But it also could take years to figure out how valuable it is, because the companies working on this don’t yet know who lived or died. On a broader level, though, while we talk a lot about finding new sources of data, it is very possible there’s some meaningful insights in old data, just waiting for the right machine learning models.
Sponsor: Bonsai
Sponsor: Bonsai
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.