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

Highlights

🌍 Sam Altman: "No One Knows What Happens Next"

Discussed at [01:36]

Sam Altman was on Theo Von's podcast discussing AI's uncertain future. When pressed about the rapid pace of development, Altman repeatedly said "no one knows what happens next" and called AI development "this weird emergent thing." On the competitive landscape, he confirmed there's definitely a race happening, comparing it to the old "megahertz race" in computer chips where companies competed on benchmarks that eventually became irrelevant.

Kyle's take: The honesty is refreshing - at least he's admitting nobody has a clue what's coming. Even himself!

But here's the scary bit: you can't become CEO of a company like this without having an ego the size of a planet. That's just how these mega-corp leaders are wired - same with Steve Jobs, same with any of them. The fact that it relies on one individual is problematic.

We're in this weird place where gigantic corporations are the only ones who can afford to build these models - we're talking billions moving into trillions. So we've got to pick which is the least bad option. That's terrifying for humanity.

Source: Theo Von Podcast

🔥 The Zero-Click Internet is Killing Websites

Discussed at [17:08]

New data shows websites that used to rank first on Google can lose 79% of their traffic if they appear below an AI overview. Google's own AI summaries are destroying the click-through rates that built the entire internet economy. It's an existential threat to any business that relied on SEO for the last two decades. It’s ALSO an existential threat to Google. They are cutting the branch they are sitting on.

Kyle's take: This is the slow-motion car crash I've been banging on about. Google makes $200 billion annually from search ads at 75% profit margins - it makes up 80% of their entire business model.

But AI overviews mean people don't click through to websites anymore. No clicks means no ad revenue for Google, but also no visitors for businesses. It's like Google is cannibalising its own golden goose whilst trying to build a new one.

Problem is, no AI company is generating anywhere near $200 billion yet. OpenAI’s revenues doubled to $10bn this year - not even close. Google may be cooked if they can’t pivot out of this.

Source: Guardian

🎬 YouTube's Mixed Up AI Policy

Discussed at [35:39]

Strange few weeks for YouTube: they curtailed AI content monetisation on July 15th, saying they want "authentic” human-made content only. Then a week later, they launched Veo3 integration directly into Youtube Short.

You can now upload a photo and automatically turn it into a video. Classic Google - the left hand doesn't know what the right hand is doing.

Kyle's take: This is peak Google dysfunction. One team says "no more AI slop, it's ruining our platform," whilst another team literally builds AI slop generators into the platform.

Shows how massive corporations struggle with coherent strategy when they're trying to have their cake and eat it. YouTube wants to protect content quality whilst also wanting to dominate AI video creation. Pick a lane, Google.

Source: YouTube Official Blog

🎲 AI Can't Do Random Numbers (Just Like Us!)

Discussed at [42:16]

Brilliant discovery from the community: ask ChatGPT, Claude, or Gemini for a random number between 1-50, and they almost always give you 27, 37, or 42.

But turns out when you ask humans the same question, they too will often answer 37. These models are literally emulating our psychological biases about what "seems random."

Kyle's take: This is fascinating, not terrifying. We humans are rubbish at true randomness - 37 is the number most humans pick because it "feels" random whilst avoiding “less random” choices like 25 or 50.

The fact that AI does this too shows it's learned our patterns, including our quirks and flaws. It's like having a digital mirror of human psychology.

As a fun aside the number 42 appearing frequently too is brilliant - that's the joke answer from Hitchhiker's Guide to the Galaxy sneaking into AI training data. These models have accidentally absorbed our sense of humour.

Check out this YT short on the issue

Member Question from Back in Shape: "Are the new ChatGPT agents any good?"

Kyle's response: If you've been using Manus for months like I have, they're not particularly impressive - about the same capability. But for most people, this will be crazy effective because it's built directly into ChatGPT. The real value here is accessibility, not capability. Having agents that can break down complex tasks into smaller chunks whilst you do other things - that's the leverage we've been promised with AI.

This question was discussed at [10:16] during the live session.

Want to submit a question? Drop it below this video and I'll cover it in a future live.

Want the full unfiltered discussion? Join me tomorrow for the daily AI news live stream where we dig into the stories and you can ask questions directly.

Streaming on YouTube and TikTok (follow and turn on notifications for Live Notification).