Skip to content

UL NO. 435: Making New Things is Post-AI Safety

Jason Haddix's AI Course, Microsoft Recall analysis, exercise erasing trauma, AI and the jobs problem…

Jason Haddix's AI Course, Microsoft Recall analysis, exercise erasing trauma, AI and the jobs problem…  ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌

UL NO. 435: Making New Things is Post-AI Safety

Jason Haddix's AI Course, Microsoft Recall analysis, exercise erasing trauma, AI and the jobs problem…

👉 Continue reading online to avoid the email cutoff issue 👈

SECURITY | AI | MEANING :: Unsupervised Learning is my continuous stream of original ideas, analysis, tooling, and mental models designed to help humans thrive in an AI-powered world.

TOC

  • NOTES
  • MY WORK
  • SECURITY
  • TECHNOLOGY
  • HUMANS
  • IDEAS & ANALYSIS
  • RECOMMENDATION OF THE WEEK
  • APHORISM OF THE WEEK

NOTES

Hey there,

Lots of good stuff this week:

  • My definition of prompt engineering JUMP
  • My analysis of AI’s impact on the job market JUMP
  • I’m working on a number of new talks. I didn’t think I’d be doing this. I planned on having like 2-3 for the whole year, and just doing them different places. But I like this problem. It’s better than the opposite of having no new ideas.
  • I published a new official_pattern_template in Fabric. It’s basically all my latest format and instruction tricks in one place, and it’ll be the new model I apply to all my prompts. And I’ll keep it updated as the state-of-the-art advances and empirical testing becomes available. MORE
  • I did my buddy Jason Haddix’s RED BLUE PURPLE AI class last week, and it was phenomenal. Jason is the best teacher I know and going to his classes are a unique combination of learning, hanging with friends, and bettering yourself—all in one. Absolutely loved both days of content. When it becomes available again, you should sign up.

Ok, let’s get to it…

MY WORK

Check out this new sponsored conversation I had with Abhishek Agarwal, co-founder and CEO of Material Security. We talk about:

  • why product managers are such good co-founders
  • the need for customized security measures for different organizations
  • the role of AI in detecting email threats
  • the importance of single-tenant environments for sensitive customers
  • and the potential risk of default settings in productivity suites like Google Workspace

Abhishek seriously built this service the same way I would have, and I absolutely loved the conversation.

YouTube video by Unsupervised Learning

A Conversation with with Abhishek Agrawal from Material Security

🎙️ My new piece on getting the perfect sound from your microphone, including a bunch of audio tips and analysis of the sounds of top podcasters.

Analysis of Mics and Mic Sounds Used by Podcasters

The difference between different mics, their sounds, post-production, desired sounds, and other podcast-related microphone information

danielmiessler.com/p/podcast-audio

This one is political, but centrist. I think you’ll like it if you’re not extreme on either end, but if you don’t like political content—skip it. 😀

The Left's Brexit

The right hobbled the UK with Brexit, and now the left in US has done the same with Stormy Daniels

danielmiessler.com/p/left-brexit

SECURITY

Check Point Vuln
Check Point swiftly released hotfixes for a VPN zero-day, CVE-2024-24919, which is being exploited to infiltrate networks.

Massive Botnet Bust: $5.9 Billion Stolen
The US and international agencies dismantled a colossal botnet responsible for stealing $5.9 billion. This botnet, known as 911 S5, infiltrated over 19 million IP addresses across nearly 200 countries. MORE

AI Against Jamming: Ukraine's Drone Solution
Ukraine's special forces developed Eagle Eyes, a software that lets drones navigate using AI and machine vision, bypassing Russian jamming. Eagle Eyes uses AI to compare live video with pre-collected maps, enabling drones to operate independently of GPS or operator signals. MORE

💡This is the exact tech from Daniel Suarez’ Kill Decision. Basically, offline drones that didn’t need a connection to any network or signal to function.

Sponsor

Enhance Enterprise Security: Trust Every Device with 1Password!

When you go through airport security, there's one line where the TSA agent checks your ID, and another line where a machine scans your bag. The same thing happens in enterprise security, but instead of passengers and luggage, it's end users and their devices.

These days, most companies are pretty good at the first part of the equation, where they check user identity. But user devices can roll right through authentication without getting inspected at all. In fact, 47% of companies allow unmanaged, untrusted devices to access their data. That means an employee can log in from a laptop that has its firewall turned off and hasn't been updated in six months. Or worse, that laptop might belong to a bad actor using employee credentials.

1Password finally solves the device trust problem. 1Password ensures that no device can log into your Okta-protected apps unless it passes your security checks. Plus, you can use 1Password on devices without MDM, like your Linux fleet, contractor devices, and every BYOD phone and laptop in your company.

Visit 1Password.com/unsupervisedlearning to watch a demo and see how it works.

1Password.com/unsupervisedlearning

Watch a Demo

Smart Home Tech in Warfare
Home Assistant is being repurposed to alert people of incoming missile and drone attacks in Ukraine. MORE

Data Deletion Services
Incogni is another service like DeleteMe that removes your info from over 170 data brokers. MORE

Sponsor

Build Trust and Accelerate Growth with Vanta

Whether you’re starting or scaling your security program, Vanta helps you automate compliance for SOC 2, ISO 27001, and more.

Streamline security reviews by automating questionnaires and demonstrating your security posture with a customer-facing Trust Center.

Over 7,000 companies like Atlassian, Flo Health, and Quora use Vanta to manage risk and prove security in real time.

vanta.com/ul

Get $1,000 off Vanta

Ticketmaster Breach
Ticketmaster got hit by a group called ShinyHunters, and it looks like around 560 million people’s data has been stolen. MORE

A Few Spread the Most Fake News
In 2020, a tiny group of 'supersharers', mainly older Republican women, spread 80% of the misinformation online. The studies reveal that small, persistent groups can have a significant impact on propaganda and misinformation, and in this case, especially on vaccine hesitancy. MORE

NSA's Easy Hack Protection Tip
The NSA released a guide to keeping your phone from being hacked. One of its top recommendations was simply to reboot your phone once a week. MORE | THE REPORT PDF

👉 Continue reading online to avoid the email cutoff issue 👈

TECHNOLOGY

Speed King: Groq's Llama 3 8B
Groq's Llama 3 8B model exceeded 1,200 tokens per second. I’m telling you, this thing is FAST. It seriously feels fake. It’s like the result is pre-run and stored, but it’s actually running in real time. MORE | GO CHECK IT OUT

💡So here’s the question: why doesn’t Google or OpenAI or Microsoft just offer this guy $1 billion in cash for this technology? Or has that happened already? This is one of the most insane stories in AI right now.

One guy basically has magic beans (proprietary chip tech) that lets him run AI multiple times faster than anyone else in the world.

OpenAI's Apple Deal: A Game Changer?
Satya Nadella is reportedly freaking out over a potential OpenAI and Apple collaboration. Cool story but I just want Apple to fix Siri. If that doesn’t happen, nothing happened. MORE

Outperforming Standard Attention
Researchers have developed new attention mechanisms that actually outperform the standard multi-head attention. MORE

Recall: A Privacy Disaster?
“Stealing everything you've ever typed or viewed on your Windows PC is now a reality, thanks to a feature in Copilot+ Recall that's a privacy nightmare.”

AI wrote that, and while I think it’s true, I disagree with the take. As I talked about last week, this tech is pure magic. In 10 years, almost everyone will think computers without this functionality are worthless. MORE

💡Actually in 10 years most people will be talking and gesturing to their computers. The “computer” will mostly be monitors and/or HUD displays, and your personal AI will be the one doing the work.

So the idea of your AI forgetting work that you did, or who said what when, will be a completely asinine idea. OF COURSE your AI remembers everything! How else would it be useful?

LLMs Aren’t Just Internet Simulators Anymore
LLMs are increasingly being trained on custom, non-internet data, enabling them to exceed the limitations of "internet simulation”. MORE

Nvidia vs. Apple
Nvidia is getting ready to overtake Apple in market cap. Bonkers. MORE

💡Here’s a way to make sense of this:

Imagine that there’s a giant filter on how many people are capable of:

  • Creating a hit movie
  • Writing a book
  • Starting a business
  • Sharing their art with billions of people

In other words, imagine the current number of major builders/creators is something like .0000000000017% of our 8 billion people. (I made that number up, just work with me.)

The reason NVIDIA is rising so fast is because it’s one of the primary pieces of tech that will remove 3-5 zeroes from that number.

We’re talking about multiplying humanity’s creative output by THOUSANDS in the next decade.

That requires chips.

Thank you for coming to my TED talk about NVIDIA.

Google's Playbook for Keeping Things Simple
Google's SRE Handbook champions simplicity as a core principle for reliability. MORE

💡They need one for product management.

Kaparthy’s GPT-2
🔧 Reproducing GPT-2 (124M) in llm.c in 90 minutes for $20 | by Andrej Kaparthy | MORE

Tinygrad Updates
🔧 Tinygrad's latest update slashes Python time and ditches external dependencies. | by tinygrad | MORE

Terminal Animations That Will Blow Your Mind
🔧TTE: Terminal Text Effects brings a splash of visual flair to your terminal with animations like beams, binary paths, and even a black hole effect. MORE

👉 Continue reading online to avoid the email cutoff issue 👈

HUMANS

INR: The Spot-On Predictions Bureau
There's this little-known intelligence bureau that nailed the outcomes in Vietnam, Iraq, and Ukraine. MORE

💡I am going to build some kind of programmatic / AI way to get this group’s predictions into my field of vision.

Sleep's Economic Divide
This study says sleep inequality in the US is deeply tied to economic stress and health disparities, not just bad "sleep hygiene." MORE

Exercise Rewires Brain to Forget Trauma
Mice with PTSD-like behaviors showed significant improvement after having access to a running wheel, hinting at exercise-induced neurogenesis as a potential therapy. MORE

Taste the Difference: How Ozempic and Wegovy Could Change Eating Habits
Ozempic and Wegovy might be fine-tuning your taste buds to help you lose weight by making sweets taste sweeter. MORE

Carmack Questions Work's Worth
John Carmack dives into "Bullshit Jobs," sparking a lively debate on the value of work. Carmack's review sheds light on the often unspoken reality of perceived value versus actual productivity in modern employment. MORE

The Social Lives of the Teens Who Don't Have Phones
“Imagine navigating high school's social maze without a smartphone, missing out on chats and memes but gaining a unique perspective on life and friendships.” MORE

💡I don’t have to imagine; I grew up in the 80’s. It was glorious.

Could Eye Exercises Be the Key to Fighting Myopia?
A comprehensive review finds eye exercises might actually help in preventing and controlling myopia. MORE

Lung Cancer Breakthrough: Lorlatinib's Unprecedented Success
A new lung cancer drug, lorlatinib, has shown unprecedented success, keeping 60% of advanced-stage patients progression-free for five years. MORE

Marc Andreessen Wants You to Stay in School
Marc Andreessen's advice to a Stanford student to "stay in school" is surprisingly grounded. The real kicker is his reasoning: if you're the type to drop out, you probably wouldn't listen to advice to stay anyway. MORE

I Love My Wife. My Wife Is Dead.
Richard Feynman's undelivered letter to his late wife reveals a heart-wrenching blend of love and grief, even years after her death. MORE

👉 Continue reading online to avoid the email cutoff issue 👈

IDEAS & ANALYSIS

Here’s my new favorite way to explain prompt engineering. In short, don’t think of it like an AI thing, or a tech thing. Think of it this way instead.

Prompt Engineering is how to explain to a superior intelligence what your problem is and what you’d like to happen. If you can’t do that well, you’ll lose to people who can.

My X thread on why it’s so hard to find tech jobs right now, and how I think AI is affecting the situation:

The problem for the job market isn't that AI is happening. The problem is that it’s happening at the exact same moment that most companies are figuring out that 80% of their employees are worthless.

We need to stop expecting things to go back to the way they were.

The new reality is companies mostly hiring super ambitious, exceptional, proven people who are gods with AI.

This means most formal education becomes a waste of time and money because a degree doesn’t certify you are any of those things.

In this model, only elite schools will matter because the filtering for being exceptional will have happened just by being accepted into the school.

It’s like two separate things:

1. You’re exceptional enough to be accepted here.

2. You finished some classes.

#2 you can get anywhere. #1 is the thing that employers actually care about.

So the result will be companies hiring the top 10-20% of people in competence. This will be filtered by:

1. Elite school attendance (if young)

2. Proof of competence via something you’ve put into the world, like on your website, YouTube, as a tool, or as a company you built.

So, in hyperbolic form:

If you’re not 19 and at Harvard—or have your own projects or companies you’ve built and talked about online—you are not going to be interesting to employers.

You’ll be part of the 90% fighting for scraps.

tw profile: ᴅᴀɴɪᴇʟ ᴍɪᴇssʟᴇʀ

ᴅᴀɴɪᴇʟ ᴍɪᴇssʟᴇʀ @DanielMiessler

tw

The problem for the job market isn't that Al is happening.

The problem is that Al is happening at the exact same moment that most companies are figuring out that 80% of their employees are worthless.

We need to stop expecting things to go back to the way they were.

12:24 AM • Jun 2, 2024

51 Likes   9 Retweets

8 Replies

RECOMMENDATION OF THE WEEK

For everyone you know who is going to be thinking about their career in the future—which I guess means everyone who’s not independently wealthy—try to get them thinking in the following way:

  1. AI is going to take out most executors of knowledge work.
  2. The people who will survive are the people with ideas who are actively. building/making that thing.
  3. So the safest thing to be is a CREATOR or a BUILDER.
  4. That means founder, entrepreneur, programmer with ideas, productive and ambitious artist, etc.
  5. The trick is you have to be able to 1) have the ideas, 2) make it yourself or get the talent/AI together to have it made, and then 3) market the hell out of it.
  6. In short, the winners in this new game will be the people making new things and bringing them into the world.

APHORISM OF THE WEEK

The most beautiful people we have known are those who have known defeat, known suffering, known struggle, known loss, and have found their way out of the depths.

Elisabeth Kübler-Ross

Thank you for reading.

UL is a personal and strange combination of security, AI, tech, and lots of content about human meaning and flourishing. And because it’s so diverse, it’s harder for it to go as viral as something more niche.

So—if you know someone weird like us—please share the newsletter with them. 🫶

Share UL with someone like us…

Happy to be sharing the planet with you,