Unsupervised Learning: No. 165

Unsupervised Learning is my weekly show that provides collection, summarization, and analysis in the realms of Security, Technology, and Humans.

It’s Content Curation as a Service…

I spend between five and twenty hours a week consuming articles, books, and podcasts—so you don’t have to—and each episode is either a curated summary of what I’ve found in the past week, or a standalone essay that hopefully gives you something to think about.

Subscribe to the Newsletter or Podcast

?️ Security News

? The OpenAI team created an algorithm that can write news stories so well that they are refusing to release it due to potential use to create fake news. I get what they’re doing, but the odds of this not being co-developed by many other groups is close to zero over a span of months. Here’s an example of a fake story it wrote about national security, and it did this on its second try with just a few words of seeding by a human. Link

People are concerned that Twitter may not be actually deleting DMs when you delete them. Someone pulled their data archive from Twitter and found their own deleted DMs. It’s a good reminder that it’s a good policy to just consider anything you do online to be permanent. Link

Trend Micro and the Ponemon Institute created a Cyber Risk Index, which ranks from -10 to 10. Strangely, the lower the number the better, even though it’s called a Risk Index. And if you put that in a graph from left to right, the right side is better than the left side. I couldn’t even get into the way they built the numbers because I’m forced to assume that process was as bad as the interface. The whole purpose of a system like this is to be used by others, so how could you possibly make such bad choices on the UI? Link

Someone found an open Chinese database online that was being used to track the location of millions of Uyghurs in the country. Between this, the re-education camp, the mass-surveillance and social credit system, the nation-wide censorship firewall, and their colonization of Africa, it’s obvious they’re willing to do anything to win at this real-life game of Civiilization. The’ve become morally belligerent. Link

Drones will soon require visible license plates. Link

Switzerland is doing a bug bounty on their e-voting system. Link

The Pentagon is worried about China and Russia fielding ground-based lasers that can blind and otherwise disable U.S. satellites. Link

Advisories: Ubiquiti Device DDoS Potential, 75 Adobe Vulnerabilities, Android App Tracking, SAP HANA,

Leaks: 620 Million Records For Sale on Dark Web

⚙️ Technology News

If you refresh this website you’ll get a human face, except they’re not real humans—they’re AI generated. Importantly, it’s not a collection of pre-made images getting loaded. They’re all generated on the fly. Link

Chinese phones now make up a third of the European market, and Huawei has the top position. Link

One in six Americans wear a smartwatch. Link

??  Human News

Student debt that’s 90 or more days delinquent is now at $ 166 billion dollars—which is an all-time high. Link

Men who could do more than 40 push-ups had a 96% reduced risk of cardiovascular disease relative to men who could do less than 10, over a ten-year study. Link

Mars Rover Opportunity’s last words were, “My battery is low and it’s getting dark.” It was a little robot that was only supposed to work for 90 days, but it soldiered on for 15 years. And after not being able to raise her after many attempts, they sent her Billie Holliday’s I’ll Be Seeing You. It’s strange how I get emotional about such things, but I’ll miss her. I hope to be able to watch when we revive her once on the planet. Link

NASA and ESA are planning a mission to deflect an astroid. Link

? Ideas, Trends, & Analysis

The Rise of the Corporate Technology Ecosystem — My new essay on how corporations will soon become our universal and subscription-based providers of everything from education to healthcare. Link

One of the biggest problems right now with cybersecurity insurance is that the exceptions are too many and too large. Mondolez refused to pay on a cyberinsurance policy for NotPetya on the grounds that it was an Act of War, which gets them out of it. The problem is, with cybersecurity, these lines are extremely messy. If we can’t say who’s done what with any confidence, it gives lawyers for the insurance companies a lot of room to play with. The irony of course is that this will make cybersecurity insurance less popular, which is the opposite of what they want. Ultimately, the problem is they have no risk data on which to establish pricing, and the threats are both unknown and  constantly in flux. This will keep cybersecurity insurance in an exploratory mode for some time, but not forever. It’s my prediction that they’ll start asking for access to corporate data lakes and network taps before too long, and they’ll use algorithms to determine inherent risk based on that. Companies won’t like that, obviously, but they will want the protection so they’ll comply. This could actually end up being the way we end up with data-based risk quantification—with insurnace companies leading the way. Vendors will make claims about security, but insurance companies will be able to tell you what works and what doesn’t because they’ll be watching the wire, the data lake, and the security outcomes. Part of me hates this, because it’s movement from magic to checklists, but part of me loves it because it could be our best hope for actual effectiveness in this insane industry. Link

AR Will Spark the Next Big Tech Platform — Call It Mirrorwold. This is a Wired piece that I quite enjoyed. For anyone who read my book, it’ll sound quite familiar, but with more of a single focus on AR. Link

Are Intellectuals Suffering a Crisis of Meaning? — A fascinating discussion about the relationship between gifted people and meaning. Link

Junk Food and Betty Boop are examples of Supernormal Stimulus—where something artificial attracts the attention of our evolution-trained sensors. Basically, exaggerated versions of natural things that we can form an addiction to. The piece talks about how porn and hentai can be examples of this, but I find the junk food angle really interesting. It’s like porn for the calorie sensors in our brains, and that’s why it’s hard for regular food to compete. Link

Why you should be an engineer or political scientist or social scientist first, and then get into data science, rather than calling yourself a data scientist. I really like this analysis, and it reminds me that you’d seldom go into a job interview and say you’re good at math. The interviewer says, “Great, math doing what?”, and you say, “Oh, just math stuff, all the maths. Whatever you need with math.” Data Science is like writing, or speaking, in that way. It’s a powerful tool to put to use in doing a project, or in a field, but it doesn’t seem tremendously descriptive to just be a data scientist. I think it’s powerful to say that you can find truth from data, and can do that in multiple fields, but perhaps that should come after a description of the type of problems that interest you. Link

? Discovery

Dolphins get high on puffer fish. Link

Amazon released 5 new bare metal instance types. Link

Machine Learning Dataset Preparation Link

Kali 2019.1 Link

? Notes

First thing I thought of with the AI Face Generator was an image generator for a modern-day role-playing game. 

I finished reading The Master Switch, and it was glorious. It was all about the history of information empires, up to and including the internet today. Fascinating stuff, and it has me wanting to learn more and more history. 

Now I’m trying to finish This Will Make You Smarter

?️ Recommendations

These are my favorite four (4) podcasts. They cover security, tech, startups, and the intellectual world. You could drop all your others, just do these four, and still be in great shape. Link


“People don’t seem to realize that their view of the world is also a confessor of character.”

~ Ralph Waldo Emerson

Become a direct supporter of my content for less than a latte a month ($ 50/year) and get the Unsupervised Learning podcast and newsletter every week instead of just twice a month, plus access to the member portal that includes all member content.

Daniel Miessler