- AI with Kyle
- Posts
- AI with Kyle Daily Update 101
AI with Kyle Daily Update 101
Today in AI: ChatGPT ads are coming + What is AI?
Kyle Balmer
December 02, 2025
The skinny on what's happening in AI - straight from the previous live session:
💰 ChatGPT Ads Are Coming: Android Beta Reveals Ad Features
Leaked Android code shows ad carousels, search ads, API ads. OpenAI hired 630 Meta employees (20% of staff). The monetisation strategy becomes clear.
Kyle's take: Surprising nobody - of course they're adding adverts. That's the ONLY way they make real money.
Found in Android beta: "ads_debug_info," "search_ads_carousel," "api_ad," even something called "bazaar_content_wrapper" (shop maybe?).
They've hired 630 ex-Meta employees - not for engineering, for ADVERTISING expertise.
Remember that Google makes 70% from ads, Meta's an ad platform. ChatGPT has eyeballs, especially free users. Eyeballs = ad revenue. And OpenAI need to make some money eventually.
My speculation: free accounts get ads, paid accounts ($20-200/month) get fewer/none. If they put ads in paid accounts they better be ready for the public backlash…. unfortunately the paying customers are the most valuable to advertisers on account on them, well, having money!
This isn't about subscriptions paying the bills - never was. They’d need a sizeable % fo the earth’s population paying subscriptions to pull that off.
Once OpenAI pulls trigger, Google can move their search business into Gemini. They're all waiting for someone to blink first. I've been saying this for ages - advertising is the endgame.
Source: Android app beta leak
🤔 "That's Not Real AI!" - The Exhausting Definition Wars
Twitter erupts over cancer detection AI. People claim machine learning isn't AI, ChatGPT isn't AI. Let’s clear this up.
Kyle's take: Saw 2.5M view tweet saying specialised medical ML algorithms aren't AI, just "PR for slop generators." This is mental gymnastics.
People want AI (which they think is bad/ChatGPT) separate from useful things like cancer detection. So they claim good uses "aren't actually AI."
Jeremy Howard (created a deep learning radiology company) stepped in: "It is and always has been part of AI."
Here's the reality: AI is the field of study since 1950s. Machine learning is SUBSET of AI. Deep learning is subset of ML. ChatGPT is built from deep learning. It's all AI - been using it for decades on Amazon, Netflix, video games.
Memorise this
When people say "ChatGPT isn't real AI" they mean it's not AGI (Artificial General Intelligence). They're confusing weak AI (what we have) with strong AI (human-level). We've had AI for ages. ChatGPT just made it visible.
Source:
âš¡ AI Uses Less Energy Than Your Netflix Binge
One ChatGPT prompt = 5-10 seconds of Netflix. Data centres use 3% of world's electricity (~ half is AI). China's energy production goes vertical.
Kyle's take: Simon Willison ran the (back of the envelope) numbers: one ChatGPT prompt equals 5-10 seconds of Netflix streaming. So an hour show = 360 prompts. That's a LOT of prompting.
Yes, data centers jumped from 1.5% to 3% of global electricity in two years, thanks to AI. But 97% goes elsewhere. People screaming about AI energy use while binge-watching Netflix for hours are hypocrites.
That said we will need more energy. That’s sort of how we progress as a civilisation - especially on the Kardeshev scale.
China's energy production is VERTICAL - 10,000+ terawatt hours vs US flat at 4,000. They're coating deserts in solar panels. Xi Jinping called AI strategic priority years ago, built infrastructure. Meanwhile we're arguing about whether our grid can handle it.
Visual Capitalist
Here's the twist: AI companies NEED cheap energy to win, so they're investing in nuclear and renewables. We've accidentally aligned profit motive with environmental goals. They might fix climate change not from goodness but from greed. Ironic.
Source: Simon Willison analysis (blog)
🎯 AGI vs ASI: Why the Next Step Terrifies Everyone
Another misconception in AI is the difference between
narrow AI vs general vs super intelligence. From AGI to ASI might take weeks, not decades.
Kyle's take: People confuse terms constantly. Narrow AI (ANI) = specific tasks, can't self-expand. Everything today is ANI - ChatGPT, self-driving cars. Can use tools but can't do everything humans do.
They are like a horses in a decathalon. Can run and jump super fast but terrible at the javelin throw.
AGI = human-level, broad tasks, reasoning, self-improvement. Whoever gets there first owns the economy - deploy million AI workers. This is why US vs China matters. They are both sprinting towards this reality because the idea of the other reaching it first is untenable. Costs be damned.
ASI = superintelligence, combined intelligence of all humans. We can't comprehend it - like chickens looking at farmers wondering "am I friend or food?"
The terrifying bit: Geoffrey Hinton says AGI to ASI might happen in WEEKS because AGI can improve itself. Once it works out recursive self-improvement that’s the ballgame. We've spent decades getting to AGI. AGI might solve ASI overnight.
🎬 The Thinking Game: DeepMind Documentary Free on YouTube
Five years of filming DeepMind, including AlphaFold Nobel Prize moment. From makers of AlphaGo documentary. Now free on Youtube!
Kyle's take: Remember AlphaGo beating Lee Sedol at Go in 2016? That documentary showed triumph then immediate melancholy - "oh God, we just took this from humans."
This new film follows Demis Hassabis through AlphaFold solving 50-year biology problem (won Nobel Prize). Free on YouTube for AlphaFold's 5th anniversary.
Source: Google DeepMind YouTube channel
Member Questions:
Kyle's response: Like any tool - it sorta depends how you use it. Self-directed people use AI to learn, expand horizons, build impossible things. I've coded things I never could before. But most people aren't self-directed. They need dragging through education. For them, AI is the ultimate shortcut. Saw photo from Harvard library - 50 laptop screens, ALL on ChatGPT. They could be reinforcing learning or cheating. If I was 18, I'd be cheating. The process of manually writing essays builds critical thinking. Skip it with AI, you never develop those muscles. AI magnifies what you already are - curious people get smarter, lazy people get lazier. Society question: how many are self-directed vs want easy life? I fear it's mostly the latter.
Kyle's response: TikTok Live Studio for Mac (finally released!), Ecamm for YouTube/Twitter simultaneously. Mac Mini M4 Pro with 24GB memory - beefy enough for 4K 60fps recording. Sony EV10 ii camera, Shure MV7B mic. About £3000 setup that I figured out with ChatGPT instead of paying someone!
"I'm an indie filmmaker using AI for parts I can't do like social media" Keith commented at [00:36:55]
Kyle's response: This is the way! Use AI for admin, reporting, social posts - the stuff that gets in way of real work. For me it's bookkeeping, taxes, admin. Offload drudgery to AI, spend human energy on creative parts - writing, filming, editing, meeting people, networking. Parts requiring actual humanity. Unfortunately many creatives won't touch AI for any aspect, which is shortsighted.
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.