Unsupervised Learning Newsletter NO. 377

STANDARD EDITION (UPGRADE) | NO. 377 | APR 10 2023 | Subscribe | Online | Audio

Unsupervised Learning is a Security, AI, and Meaning-focused podcast that looks at how best to thrive as humans in a post-AI world. It combines original ideas, analysis, and mental models to bring not just the news, but why it matters and how to respond.

Happy Monday—I hope you're well!

I've been obsessed with langchain this past week. It's like the coolest tech in AI right now, not counting GPT-4 and Midjourney. It's basically the connective tissue for building AI applications. Think of it like the pipe "|" command in Linux. You should check it out. MORE

Meanwhile, let's get into the week!

In this episode:

🛡️ Semgrep + AI: Enhancing code security
📉 Goldman Sachs predicts AI's impact on jobs
🤖 AI's threat to newsletters & the future of content
🔍 Unmasking money and power with AI's transparency
🌉 Exploring the gap between pre-AI and post-AI worlds
📚 AI replacing tutorial webpages with assistants
🚰 The power of building pipelines over fixing things
📚 NIST AI Risk Management Framework & playbook
🚗 Tesla's video sharing privacy concerns
📸 Iranian women monitored by public cameras
🌆 Crime density stats: SF vs. Baltimore
💻 Microsoft's TaskMatrix.AI & Langchain's success
👴 Hiring older workers for their work ethic
🏭 US manufacturing resurgence & green efforts
🔭 JWST captures stunning Neptune image


MY WORK

AI's Threat to Newsletters
Most newsletters won't survive the burst of AI-generated completion we're about to see. Only one kind will likely survive. MORE

Unmasking Money and Power With AI
GPTs will magnify transparency in the world, and make it easier than ever to "follow the money". MORE

The Pre-AI World, the Post-AI World, and the Gap In-between
Can AI lead us to a post-capitalist society focused on creativity and humanity? MORE

AI is the End of Tutorial Webpages
AI assistants will replace traditional tutorials and change how we consume information. MORE

Don't Fix Things; Build Pipelines
Why pipelines are a superior mental model for problem-solving. MORE


SECURITY NEWS

Semgrep + AI
Semgrep added GPT-4 to their offering to help identify false positives and fix code. The integration allows them to auto-triage findings, fix code automatically, and provide direct assistance to developers. MORE

NIST AI Risk
NIST released its AI Risk Management Framework at the end of January, and it now has a companion playbook and video explainer. The framework is based on four primary concepts: Govern, Manage, Map, and Measure. FRAMEWORK | PLAYBOOK | EXPLAINER

Tesla Video Sharing
Ex-Tesla employees said it was possible to get access to sensitive in-car video recordings within Teslas, including 'scenes of intimacy'. Pretty much the privacy nightmare—employees passing around stuff that can cause you reputational harm. MORE
 
CISA Orders BackupExec Patch
CISA added five issues to its list of vulnerabilities abused by threat actors, and three of them were Veritas Backup Agent issues used to deploy ransomware. MORE

Iranian Women Cameras
Iran is installing cameras in public to catch and penalize women who don't wear the veil. The system starts with a warning text message informing them of the consequences. MORE

Crime Density Stats
Someone did some analysis on crimes per square foot of major cities, and San Francisco came out worse than Baltimore. MORE | MORE

TECHNOLOGY NEWS

Microsoft Taskmatrix.AI
Microsoft has introduced an AI API infrastructure called TaskMatrix.AI. It works by, "integrating foundation models with millions of existing models and system APIs, resulting in a “super-AI” that can perform various digital and physical tasks." MORE | PAPER

Mild Layoffs?
Crunchbase says the massive layoffs of over 200K people in the last 14 months only add up to around 8% of the people hired after the pandemic. MORE 

Computer Shipment Declines
Computer sales are way down, with Mac shipments down 22% and most PC makers slumping over 40%. Related to that, Foxconn also dropped 21% in March, and they expect further decreases. MORE 

Langchain Raise
Langchain has raised a $10 million seed round led by Benchmark. It's the tech powering a ton of the advanced GPT-based apps you're seeing right now, and I think they're crushing it. MORE 


HUMAN NEWS

Goldman AI Loss
Goldman Sachs says 300 million jobs will be lost or degraded by AI. Sounds directionally correct, but the real questions are 1) how many will also be created, and 2) how quickly all these things will happen, and 3) how people and economies will respond. MORE | PAPER

NATO Finland
Finland is now part of NATO. MORE 

SF Murder
A well-known tech entrepreneur, Bob Lee, was stabbed and killed in San Francisco last week. He had moved out of the city for safety reasons. As a former SF resident, this resonates strongly with me. I really do feel like the city is lawless, and as a liberal, this angers me. I was told by people growing up that you don't want to live in liberal cities because they ruin everything. I didn't believe them. I still don't think that's the case everywhere, but it sure seems true for San Francisco. Too much of it looks like a set for The Last of Us, and I feel uneasy whenever I'm there now. MORE 

Hiring Older
Bosses need hard workers, so they're hiring older people who have a different work ethic. MORE 

America is Back in Manufacturing
The US spent $108 billion on manufacturing-focused construction in 2022, which is more than schools, healthcare facilities, or office buildings. The trend is tied to green efforts and a desire to secure our supply chains. MORE

JWST Neptune
The JWST telescope just caught a wicked photo of Neptune, showing its moons and its massive ice ring system. MORE


IDEAS & ANALYSIS

Mind the Font
If you're blogging on a platform where the font looks the same no matter who's writing—i.e., your content looks identical to everyone else's—that means it's their platform, not yours. When you read the New York Times, you're not reading authors, you're reading The Times. Maybe that's ok if you're building a career as a writer and you're picking big names to establish credibility, like writing for the New Yorker, then The Atlantic, so you can make it to the New York Times. That makes sense. Or at least it did like a year ago (not sure about now). But as a new blogger that's not your situation. Your goal should be broadcasting as yourself as purely as possible. And that means controlling the look and feel of your content. It should look and feel different than other people's stuff. Blogging on Medium or Substack is like showing up to do a talk wearing a company uniform, covered in their logos. Again, not bad if it's the best brand in the world, but probably not worth it if it's not.

The End of Googling
We need a new word for Googling something. I've probably switched to GPT for 85% of my searches in the last month or so. Paul Graham said his 10-year-old doesn't Google things anymore. He either uses GPT or Pokedex. Makes sense. So what's the term? GPT'd? Too long. AI'd? Cumbersome. I wonder if it'll just be called "searching", and the Kleenex naming advantage Google had will just fall off. What are your guesses for the new verb?


NOTES

I've had my life elevated by a couple of utilities in the last week. 1) Espanso. It's a Rust-based, YAML-configured text expansion system. It lets you run shell commands as well so I've integrated it with my AI APIs for doing quick little shortcuts like ;essay, ;correct, ;proofread, ;analyze-incident, etc. to get instant analysis within any text location, including in my terminal. 2) MacGPT. It's a little utility that lets you call GPT directly from your Mac. You just call up the shortcut and a little window pops up and you type in your question. For whatever reason it's lighting fast. The combination of these two things have accelerated my life greatly in the last couple of weeks.

My buddy Matt Jay has a new newsletter, and he's on the sixth episode right now. It's called Vulnerable U, and its focus is on how we can use vulnerabilities, and vulnerability, as a catalyst for growth. Matt's content is spectacular, and I recommend you check it out. MORE 

DISCOVERY

⚒️ The Lex Fridman Podcast ChatGPT Plugin
ChatGPT now has access to every episode of Lex's podcast. MORE | BY RILEY TOMASEK

⚒️ A GPT-powered Poem Clock
Someone made a clock out of poems, using GPT to create the rhymes. MORE | BY MATT WEBB

Why I Blog MORE | BY DANNY GUO

Langchain as a ChatGPT Plugin MORE

🔥 OpenAPIEndpointChain — Wraps a single API endpoint in a natural language interface. MORE

Everything is a Practice MORE | BY DANNY GUO

How to create a podcast search tool using Lanchain Tools. MORE

Video is the next target for Generative AI, and things are already getting crazy. MORE

Midjourney has banned all images of Chinese President Xi Jinping. MORE

RECOMMENDATION OF THE WEEK

1. Espanso for text expansion.

2. MacGPT for immediate access to GPT-4.

3. Constantly be thinking about the services you provide and where they sit in the hierarchy of ideas. What about what you do can be replaced easily, and what about what you do is resilient? Are you the SME? Are you creating AI tools? Or are you the polymath glue that solves problems with AI?

 APHORISM OF THE WEEK

"To be yourself in a world that is constantly trying to make you something else is the greatest accomplishment."

Ralph Waldo Emerson

Earn rewards for sharing UL…

You can now earn rewards by sharing UL with others! Share to earn:

- 5 Shares -> A PDF on How to Thrive in a World Full of AI

- 10 Shares -> 25% Off UL Membership

- 20 Shares -> 50% Off UL Membership

- 50 Referrals -> A Private 30-minute Mentoring Session With Daniel

- BONUS: Plus, each referral is an entry to win a pair of AirPods Pro 2 on May 31st!

- Congrats to Charles Blas for winning the first AirPods Pro 2 giveaway on February 28th!