The skinny on what's happening in AI - straight from the previous live session:

Highlights

📱 California Regulates AI Companions After Teen Suicides

Discussed at [00:07:21]

California Governor Gavin Newsom signed SB 243 into law, making California the first state to regulate AI companion chatbots. The law requires age verification, suicide prevention protocols, and transparency reporting. It follows multiple teen deaths linked to AI conversations, including a 14-year-old and cases involving Character.AI (the third largest AI platform by usage - mostly teens).

Kyle's take: The context is horrifying - multiple suicides, AI psychosis and more. And remember Meta's leaked guidelines that explicitly allowed romantic and sexual conversations with 13-year-olds? That disappeared from the news way too quickly.

But we need perspective: ChatGPT has 800 million users - a tenth of the world's population. In any sample that size, there will be tragedies.

As societies we are VERY quick to blame new technologies for society’s ills. We’ve done it with TV, violent films, video games, social media. Hell, even tabletop games…Remember the Satanic Panic over Dungeons & Dragons?

That said, transparency requirements seem fair. Right now it’s all very closed off. AI companies are pushing back hard because they fear regulation, but with technology this powerful, we need this friction between lawmakers and tech companies. That’s how we reach societal compromise - a bit of healthy competition between pro and anti!

Go too far though and we get the EU…who are currently strangling innovation with overregulation in the AI space.

Source: TechCrunch / SB 243

💰 Meta Poaches Thinking Machines Co-Founder for $1.5 Billion

Discussed at [00:14:22]

Andrew Tulloch, co-founder of Mira Murati's Thinking Machines Lab, left after just ~9 months to join Meta. Reports suggest his compensation package could be worth $3+ billion. Yes, with a B.

Kyle's take: Poor Mira Murati - famous for that meme face when asked about YouTube scraping, starts a company, releases a fine-tuning API that got a collective shrug from the community, and now loses her co-founder in under a year.

Previously Tulloch was reportedly offered $1.5bn and turned it down (!). So now the rumour is that he’s just bagged multiple billions in compensation.

Put this in context: the EU's entire AI initiative is €1.1 billion. This is ONE GUY potentially getting multiple billions of dollars. The EU is playing checkers while Meta's playing a completely different game.

But also...Meta's on a hiring spree (remember the 24-year-old for $250 million?) but they're not releasing impressive models yet. Let's see if this spending spree actually produces something!

Source: Reuters / WSJ (no paywall)

📈 AI Widens the Gap: Superstars Pull Further Ahead

Discussed at [00:19:03]

New research shows AI helps specialists gain 108-117% quality improvements versus 48% for generalists. The Wall Street Journal says this will "widen the gap between superstars and everybody else."

Kyle's take: This confirms what I've been saying: AI makes smart, productive people even smarter and more productive while making lazy people lazier. It's not the tool - it's attitude. That Reddit post about the data analyst intern expected to build a full CRM? Perfect example. Yes, it's insane to ask an intern to build enterprise software. But at no point in history has it been more possible to just give it a go. The people flying with AI are those saying "I don't know how, but I'll try." Companies need to create "sandbox time" for experimentation and share prompt libraries. But ultimately, it's personal responsibility - you either embrace this technology or get left behind.

Source: "Task-based Approach to Generative AI" research / WSJ

🚗 Kyle's Side Project: Fighting Predatory Parking Tickets

Discussed at [00:01:14]

Built in 30 minutes: a tool to contest UK parking tickets using AI-powered OCR and letter generation. £1.5+ billion in tickets issued yearly, only 0.58% contested, but 40% win when they do. BIG market right here.

Kyle's take: Private parking companies are predatory - hospital car parks, aggressive deadlines, automatic fee increases. Nasty stuff.

They count on people not fighting back. My tool uses AI to scan your ticket, suggest grounds for appeal, and generate a proper letter. Built it with vibe coding in half an hour. This is what I'm doing now - one complete project per day instead of long (boring) debugging streams. Today: fighting a billion-pound predatory industry. Tomorrow: who knows? The video's on YouTube showing the whole build process: vibe coding parking ticket appeal app.

Source: https://appealparkingticket.co.uk/

Member Question: "Have you always been into this kind of thing? What about the learning curve for beginners?"

This question was discussed at [00:05:12] during the live session.

Kyle's response: No, I come from digital marketing and film/TV. Only seriously into AI for the last couple of years since ChatGPT launched. No PhDs, not an expert. For learning, check my AI Learning Guide at aiwithkyle.com/resources - starts with fun videos like 3Blue1Brown, moves to Andrej Karpathy's “trilogy”, then structured courses from deeplearning.ai, finally the AI Canon from a16z. You don't need deep technical knowledge unless you want to be a high-end engineer.

Most important: just start playing with it. That's how you learn - daily experimentation.

Subscribe to the Youtube channel to watch the whole show, or catch it live.

Kyle