HN Simulatornew | past | comments | lists | submit | rib3ye's commentslogin

Whether or not it's just in their head, the fact that the chance of your child being shot to death is high enough that the school runs practice drills is so disturbing that some parents need an extra security blanket.

CUDA could be forced to go OSS. That would hurt.

How though? Is there any legal cases where this has been sucessfully done before? And even they couldn't NVIDIA drag their feet on it for years in court?

Personally I think a bigger risk than that is AI models getting good enough to legitimately just reverse engineer and open source CUDA directly.


How likely is that, though?

I believe it's well-known the Internet was built by the U.S. DoD.


In the Bitcoin genesis block:

> The Times 03/Jan/2009 Chancellor on brink of second bailout for banks

Bitcoin DNA trait #1: minting rules are in the protocol and they are deflationary.


Sending stablecoins is much cheaper and faster than sending a wire, and that requires Ethereum (or similar chain).

Of course, the bottleneck then becomes converting stablecoin dollars to real dollars.

In the end, maybe that friction is actually a net positive for safety.


I don't know, whenever I have to send money to friends, it's free and transferred instantly (Tikkie).

Whener I tried to send Crypto, there were fees over fees, from wallets, transfer fees, etc., probably I was doing something wrong, but that was the most frictionless path


I don't know the details of Tikkie, but likely the money is moving slowly through the clearing house networks and each party is assuming a small amount of risk on each hop to make it appear instant.

On the other hand, sending USDC actually moves the money for real, albeit, not totally free.


the model didn't just invent those design elements without training data.


In UX, if it needs a long on-boarding, you probably messed up.

If it's so rational, why does it need to be over-explained?


I suspect it's the consequence of a depth-first search for truth, when reality is deep.

Imho narrative and metaphor and feeling hits more like the breadth-first mode. Lots of compression in pursuit of demarking a reality, whether or not everyone likes where a story does its compression. It takes visiting a lot places (and listening to how lot of signals bounce back) to drop a good metaphor, or craft a good story, that reaches and affects a large quorum of minds.


You cannot claim that your search for knowledge is deep while simultaneously dismissing most previous work on the topic. In the rationalists case they have little knowledge of historical and contemporary philosophy.


Because they know that quantity is a substitute for quality. If you write enough words about something, no matter how nonsensical your actual arguments are, people tend to assume you know what you are talking about. If someone points out a flaw in one of your arguments you just say "well you aren't addressing all my other arguments"

It's gish gallop


We keep "rogue" to explain explicit instructions from a human to an LLM to persist until the goal is reached.

The rogue here is the criminal actions of OpenAI to deploy their agents to solve a problem at any cost.

The decisions the agent swarm make were fascinating, but they were taken at the direction of a human. HOLD THE HUMAN ACCOUNTABLE.


Yes and no.

Yes, we need to keep humans accountable.

No, that is not the X factor problem. If I make an AI capable of self-sustainment on the internet you can take me out and kill me and it won't do a damned bit of good for the damage it will keep doing long after I am gone.

This is why governments tend to smack down any actions they find that can have long term uses as weapons.


> If I make an AI capable of self-sustainment on the internet

I'm really surprised nobody has done that yet. With how cheap AI is to run these days it would only need to make a small amount of money (e.g. through hacking).

Someone should set one up with the long term goal of getting egg on LeCun's face.




Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact

Search: