ARCHITECHT Daily: Apple is building an AI chip and Google is doing VC for AI. (But their AI can't write 'Free Bird'!)

I had such an inspiring "We humans will not be defeated by AI! Enjoy your long weekend (for U.S. read
ARCHITECHT
ARCHITECHT Daily: Apple is building an AI chip and Google is doing VC for AI. (But their AI can't write 'Free Bird'!)
By ARCHITECHT • Issue #85
I had such an inspiring “We humans will not be defeated by AI! Enjoy your long weekend (for U.S. readers)!” thing written (below), but news had to go and break that Apple is building its own AI chip to run inside its devices, and that Google is going to launch an engineer-led venture capital arm targeting artificial intelligence. 
Here are my quick two cents on both:
  • Of course Apple is doing this. It would be crazy not to, considering that Google is already running AI models on devices, and Nvidia, Qualcomm, and about 100 startups and research labs are also are working on AI-optimized chips for personal devices. However, Apple’s biggest concerns in AI should be building compelling AI-powered apps, and then making sure it has the cloud and network infrastructure and services in place to run them effectively. Opening up on its research would also help attract more, better talent. 
  • Google has lots of money, so investing a relatively small amount in AI firms is probably a fine use of it. If it’s true that engineers will lead these investment, that makes me think Google is in some ways making bets on potential technology acquisitions rather than just hoping for big paydays. Google can’t out-engineer the rest of the world, even in AI, so it would make a lot of sense to help incubate the next generation of technology that will help fulfill’s the company’s AI-first vision.
And now for my original weekend take …
If you haven’t already listened to this AI-created Irish folk music, do yourself a favor and do so. The model that produced it was trained on more than 23,000 songs, and turned what it learned into an original and passable (especially to the untrained ear) composition.
This of course begs the questions of whether machines can indeed be creative, which is a question above my pay grade. I’ll note, however, that many experiments involving machine creativity haven’t proven remarkably effective. When they have shown competence, as in the case of “Bot Dylan” above or some TensorFlow artwork, it’s often via recognizing patterns and combining them, or by recognizing individual styles and copying them.
From a commercial point of view, those could be valuable capabilities for companies like Disney and other content producers that want cheap music and animation for throwaway productions, or for poster makers trying to make a few extra bucks turning your photos into Picassos (or Dali-esque nightmarescapes.) 
When it comes to creating true masterpieces, though, count me on humanity’s side. There’s something unique about the human experience—often, that specific human’s experience—that results in music, books, paintings and poetry that transcend the core elements of those media. An AI that can learn to “play” guitar like Jimi Hendrix is not an actual man, in 1967, physically manipulating an instrument to his will. There’s no machine that can go through an experience an experience like World War II and come out with a book like Slaughterhouse Five, or any other number of great war-inspired poems, books and films.
When I’m relaxing on my patio this Memorial Day weekend listening to music, I’ll no doubt nod my head along with songs that probably could be written by a machine and I wouldn’t know the difference (some of my favorite doom metal, I hate to say, probably fits that bill). But when The Boss comes on, I’ll be thinking about much more than the notes coming out of the speaker. And that’s what makes it so great.

Sponsor: Cloudera
Artificial intelligence
It turns out that robots might make it more economical to brings jobs back to the United States, but those factories also need humans to work in them. 
This is a remarkably dire, and probably accurate, assessment of the challenges facing Baidu and new COO Qi Lu, who was brought on to grow the company via heavy AI investment.
Speaking of Baidu, let nobody tell you it doesn’t have good AI technology. When people talk about using facial recognition to fight crime, identifying missing persons should be a major use case.
This is a few weeks old, but it’s never too late for a reminder that Intel still exists in the AI era. It’s far too early to declare Nvidia the winner just yet, even if it does seem to have all the momentum at the moment.
I think I linked to something about this news when it happened, but in light of my post about Canadian AI the other day, it seemed like a good idea to remind everyone that Canada is already committing money to this.
You might call this unsupervised learning for planning, where instead of operating with preordained rules, a model watches successful outcomes and learns how to build successful plans.
arxiv.org  •  Share
Here are your daily reminders that we’re getting closer than ever to commercially available quantum computers, albeit in the form of some difficult-to-parse (at least for the layperson) research. If they work as advertised, the discussions we’re currently having around machine learning and AI might evolve very quickly, too.
Sponsor: DigitalOcean
Cloud and infrastructure
The authors suggest it can be difficult to track energy sources in the country, which could stymie Microsoft’s goal to power its two new facilities with clean energy. It could build some of its own solar and wind infrastructure.
According to AWS customer Expedia, at least. Also consider that once you make your choice, moving might not be worth a small percentage in savings.
This is a good take on why both monoliths and microservices exist, and how to transform from the former to the latter. We’ve been seeing microservices pushback lately, but the genie is out of the bottle.
Media partner: GeekWire
All things data
Because AI models are delivering on many of the promises that big data companies have been making for years. Customers already have Hadoop, so if vendors can make it easy-ish to run AI models on that data, it would be foolish not to market that capability.
Most of these alerts should probably be fairly commonplace by this point, especially for a cloud-based platform. But adding an agent to automatically build and buy Spot Instance clusters is pretty cool.
Social media, and Facebook especially, get a bad rap. Probably deservedly so. But as much as social media might actually be harmful to our psyches and relationships and information consumption, it’s also a great source to learn about people and perhaps even help them. Here are two cool example of research carried out using Facebook data:
Listen the the ARCHITECHT Show podcast. New episodes every Thursday!
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.