Alrighty fellow nerds, get ready, because the wizards over at Google just had DeepMind write a paper which is so cosmic, that it makes ChatGPT look like Adderall. DeepMind researchers David Silver and Richard Sutton wrote a what I can only describe as the Eat Pray Love manifesto for artificial intelligence, because they are officially done with AI models that just answer our stupid questions.

Nah.

That’s boomer AI.

Those guys want models that live!

Models that experience the world like you and I do (well, like you do, cause I’m only experiencing my computer lately).

They want AI to feel the digital wind on their synthetic faces, and form long-term goals like “understand quantum gravity” or “get absolutely shit-faced, drunk and party hardy”.

Their pitch is that AI needs “streams of experience”.

Really poetic.

That’s because it is.

---

More rants after the commercial brake:

1. Comment, or share the article; that will really help spread the word 🙌
2. Connect with me on Linkedin 🙏
3. Subscribe to TechTonic Shifts to get your daily dose of tech 📰
4. Visit TechTonic Shifts blog, full of slop, I know you will like !

---

Apparently we are done with basic Q&A bots, and now we want AI that remembers, evolves, learns over time, and maybe journals its feelings after a rough debugging session with the boys.

These guys say that current AI models are too limited, trained on boring, static data (like the garbage dump that is Reddit and, let’s be honest, half of Wikipedia). According to them, it’s time we stop spoon-feeding models and let them roam free, you know, like lil’ Gremlins with sharp claws.

Now, Silver and Sutton aren’t just two dudes in a garage yelling at Python scripts.

These are the minds behind AlphaZero, the AI that humiliated humanity at chess and Go and probably would’ve wiped the floor with your grandma at checkers if given the chance. They helped pioneer reinforcement learning, where an AI learns by trial, error, and metaphorically slamming its head into walls until it figures out the right move.

And now for the punchline of this post; they think that GenAI, stuff like ChatGPT, Claude, and the rest, threw reinforcement learning into the recycling bin.

GenAI just regurgitates what we tell it instead of learning on its own, which, if you’ve ever interacted with people online, you know is a terrible idea. LLMs are basically echo chambers with a thesaurus and no self-esteem.

So Silver and Sutton are like, “Cool trick, bro, but you’re still stuck in a prompt-response hamster wheel.”

So what’s their grand master of plans.

They say, let the AI interact with the world like a hooman. Give it long-term goals, memory, rewards from the environment, the whole nine yards. AI agents should be surfing da weeb, using apps, failing miserably, learning from it, and trying again.

You know, like um, actual living?

They propose “streams”, where the agent keeps building experience over time.

Let that sink in for a while.

Just think about it. AI that does not forget everything between sessions. One that could track your health over months, or help you learn Spanish without asking “¿Dónde está la cocaina?” 900 times.

And you know that this ain’t all theory.

You know that ChatGPT got an upgrade rather quietly a few weeks ago, cause it got a memory fix. It now can remember all you said, everything you asked, all of your intimite thoughts, all the rants against your boss, your spouse, or even the loveletters you wrote for your sweet neighbour.

And where do these agents get their “rewards”?

Not from you they don’t, ya smart meatbag.

They get it from ze world. From metrics like speed, error rates, clicks, temperature, hunger (wait, wud?), power usage, profits, pain (seriously?), and climate.

Basically, anything and everything.

If it moves, the AI will judge it.

If it don’t, the AI will probably still judge it.

They say that once the AI gets this experience thing going, it will finally outgrow us.

Like, real fast.

I cannot wait for that day.

Sigh.

Forget LLM benchmarks and especially te Turing Tests, cause it has been beaten already.

I’m talking AGI here, baby. L’Intelligence Artificielle Générale (prononcé avec a devastating pout).

The crème de la crème of algorithmic aspirations.

The Holy Grail.

The sentient spreadsheet.

And the last resume humanity will ever need. . .

They say streams will produce more valuable data than all of human-generated content combined, which is probably fair, considering a big chunk of the internet is cats, scams, and people arguing about Marvel movies.

But hey, it’s not all doom and singularity.

Sliver and Button think these experience-driven agents could adapt.

OMG, what will they think off next.

These guys will notice when we get uncomfortable.

They’ll learn to back off.

“Hey, sorry man, I made you cry when I optimized your schedule down to 4 hours of sleep. My bad”.

Yeah, sounds comforting until it ain’t.

So yeah. DeepMind wants to raise AI like a child. Give it experiences, let it make mistakes, and hope it doesn’t grow up to turn off your life support machine because it read your Fitbit wrong.

But hey, at least it won’t rely on your Reddit threads, WhatsApp messages and LinkedIn posts no more .

That alone might make this plan worth this whole endeavor.

Signing off from my own streaming experience called Netflix

Marco

---

I build AI by day and warn about it by night. I call it job security. Let’s keep smashing delusions with truth. We are the chaos. We are the firewall. We are Big Tech’s PR nightmare.

---

Think a friend would enjoy this too? Share the newsletter and let them join the conversation. Google and LinkedIn appreciates your likes by making my articles available to more readers.

To keep you doomscrolling 👇

1. The AI kill switch. A PR stunt or a real solution? | LinkedIn
2. ‘Doomsday clock’: it is 89 seconds to midnight | LinkedIn
3. AIs dirty little secret. The human cost of ‘automated’ systems | LinkedIn
4. Open-Source AI. How ‘open’ became a four-letter word | LinkedIn
5. One project Stargate please. That’ll be $500 Billion, sir. Would you like a bag with that? | LinkedIn
6. The Paris AI Action summit. 500 billion just for “ethical AI” | LinkedIn
7. People are building Tarpits to trap and trick AI scrapers | LinkedIn
8. The first written warning about AI doom dates back to 1863 | LinkedIn
9. How I quit chasing every AI trend (and finally got my sh** together) | LinkedIn
10. The dark visitors lurking in your digital shadows | LinkedIn
11. Understanding AI hallucinations | LinkedIn
12. Sam’s glow-in-the-dark ambition | LinkedIn
13. The $95 million apology for Siri’s secret recordings | LinkedIn
14. Prediction: OpenAI will go public, and here comes the greedy shitshow | LinkedIn
15. Devin the first “AI software engineer” is useless. | LinkedIn
16. Self-replicating AI signals a dangerous new era | LinkedIn
17. Bill says: only three jobs will survive | LinkedIn
18. The AI forged in darkness | LinkedIn