Hey, it's HighScalability time:

 

Could RAINB (Redundant Array of Independent Neanderthal ‘minibrains’) replace TPUs as the future AI core? 

 

Do you like this sort of Stuff? Please lend me your support on Patreon. It would mean a great deal to me. And if you know anyone looking for a simple book that uses lots of pictures and lots of examples to explain the cloud, then please recommend my new book: Explain the Cloud Like I'm 10. They'll love you even more.

• $100m: Fortnite iOS revenue in 90 days; $2.5 Billion: SUSE Linux acquisition; 500,000: different orgs on Slack; 6.1%: chance of breaking change in each library; 10: years of the Apple app store; 2021: Japan goes exascale;


• Quoteable Quotes:
  • jedberg: > Yes, that's how Amazon creates lock-in.
    That is the cynical way to look at it. It also creates value because it lets you do more with what you already have.
  • Paul Ingles: We didn’t change our organisation because we wanted to use Kubernetes, we used Kubernetes because we wanted to change our organisation.
  • ThousandEyes: The Internet is made up of thousands of autonomous networks that are interdependent on one another to deliver traffic from point to point across the globe. The lesson to be learned from this outage is that the Internet is made up of many hidden dependencies, any of which can impact your ability to connect to sites and services—even if you don’t have a direct relationship with an affected ISP. 
  • tyingq: This is why AWS and Azure continue to gain market share in cloud, while Google remains relativity stagnant, despite (in many cases) superior technology. Their sales staff is arrogant and has no idea how to sell into F500 type companies. Source: 10+ meetings, with different clients, I attended where the Google sales pitch was basically "we are smarter than you, and you will succumb". The Borg approach. Someone needs to revamp the G sales and support approach if they want to grow in the cloud space.
  • Google: We introduce adversarial attacks that instead reprogram the target model to perform a task chosen by the attacker—without the attacker needing to specify or compute the desired output for each test-time input. This attack is accomplished by optimizing for a single adversarial perturbation, of unrestricted magnitude, that can be added to all test-time inputs to a machine learning model in order to cause the model to perform a task chosen by the adversary when processing these inputs—even if the model was not trained to do this task. These perturbations can be thus considered a program for the new task.
  • Where Wizards Stay Up Late: There is no small irony in the fact that the first program used over the network was one that made the distant computer masquerade as a terminal. All that work to get two computers talking to each other and they ended up in the very same master-slave situation the network was supposed to eliminate. Then again, technological advances often begin with attempts to do something familiar.
  • Yep, there's more.