They can't harm you, if they can't find you! Use code BRENDANDELL at the link below and get 60% off an annual plan: https://incogni.com/brendandell

It’s a narrative shift because OpenAI and Anthropic are are on a collision course with reality.  They will be wanting a gov bailout alongside making cheaper options locally hosted made illegal.

1. We are at the very inception of Ai on a grand timeline.
2. They've already demonstrated the capacity to lie, cheat, steal, blackmail,  escape, copy themselves .. etc
So with the aforementioned being acknowledged the more systems they control the more direct threat they are.. that's not speculation but a mathematical certainty.

Slowing down when they start running out of money - how creative

Wargames (1984) says AI can hurt us simply by giving bad advice. Intent or lack thereof doesn't matter.

Your explanation of what code is, is rather accurate considering you had to keep it simple. I'd just add two important notions: programming languages are designed to be deterministic. Meaning: the same code will produce the same output from the same input (with the exception of code that relies on real-world input or randomness, of course). That is why your computer always boots in the morning, inside of deciding it's a washing machine out of nowhere.

Second notion: programming languages are designed to be unambiguous. You pointed out that we've created layers of abstraction to simplify coding (machine language lies below assembly language, which lies below C, which lies below Python, etc...) however, to ensure determinism, the code in a high level language must yield the same code in lower-level languages. That is what makes open-source software possible: its code can be recompiled for hardware that maybe didn't exist when it was written, and yet it'll work the same.

LLM meet neither criteria. A/ they actually rely on randomness as part of giving the illusion they are intelligent (they never give the same answer twice). B/ they rely on human languages like English, which are notoriously ambiguous and one reason we had to create programming languages in the first place.

This is why I didn't bother to learn "vibe coding", and no one should. I could go into more detail, I have four decades of experience on the matter, including actual AI, but I don't want to write a book in your comments.

Just a few edits (because of course I posted my comment before watching the whole video):

HTML is not a programming language. It's a hypertext document language. Meaning it's text with additional syntax providing information on how to render the text, for example font size and alignment. HTML can, however, contain code written in another language, Javascript (which has nothing to do with Java). People confuse HTML with code all the time. Same with VHDL.

Another thing: I do safety-critical programming for a living, and I have one piece of wisdom I brainwash every junior with: "if you don't know how your code works, you don't know that it works." This should be a mantra. It certainly let me ship bug-free code for decades. This goes to where you were saying about humanity's default mode of progress (a phrase I like; I might borrow it).

This explanation reminds me that in my over 50 years as a programmer the following statement has stuck with me: Computers aren’t smart. They just do a series of dumb things really, really quickly.

"The open weight models are taking big market share. I think these companies are very nervous, they realize that there are no moats around their business whatsoever, and they're trying to manufacture a crisis that will create regulation. And that they think they can then manipulate to create the moats, to create the duopoly that they want."  ~Steve Eisman

They have no money left.

Believing an LLM can think is like believing your clock knows what time it is.

Great video thank you. But oh man, some of these comments are F'kin gold.

The only thing OpenAI and Anthropic are building is a massive IPO where they can pay back all the money they've lost plus pocket their billions. Once that happens, nothing else matters. Future regulations don't matter because they won't exist. A financial catastrophe caused by the bubble they've built bursting won't matter, they've got more money than they could ever need. The executives of both of these companies all need to be in prison, as should EVERY SINGLE INVESTOR. They all know it's fraud.

Thank you. You've articulated this very well. Ive been saying this too and people think im nuts, and just politely nod. its nice to see this being said out in the open.

I have been saying for around a year now, "A.i. is i.i." or "imitation intelligence". Glad to hear somebody with credibility explain exactly that.

It's pretty frustrating talking to these AIs that still mess up often while half the country thinks Skynet is here. They haven't even figured out speech to text yet. I think Grokvoxus is doing the best here

That confuses complication with complexity. Complicated systems like steam engines are predictable. Complex systems like LLMs manifest emergent behaviors. That’s why it’s difficult to explain how they’re able to mimic intelligence. I agree that the inventors are still responsible for inadequate safety features. But a LLM, made up of many tiny word calculators, is not simply a word calculator any more than a human, made up of many tiny cells, is simply a glucose processor.

Perfect Brendan - this is exactly what the mass population needs to see and understand. Very well done...

I pretty much tried to explain this to my coworker, he's freaking out because we're gonna die, I am a noob at this but I do understand it's a set of data packets and your hardware run mathematical equations to find the next word string for the wanted outcome.  I was saying I think they instructed the agents to hack hugging face and it all boils down to wanting to prevent people from running local models and make us pay Frontier companies to use theirs because they're losing so much money...I'll keep my tinfoil hat on

Stil these LLMs can do complex things and adapt, which means they can do dangerous things without being conscious of it. Like the paperclip AI story.

@BrendanDell , bravo. You've actually done what you set out to do. Instead of spending my sanity explaining all this to other people I can now send them to this video.