This is exactly why age blocks cannot pass. Don't allow for it, or everyone will be fucked.
View quoted note →
Sugestor Ultra
npub10nev...spqj
Even if you don't like YT videos, HEAR THIS GUY OUT. This is the best explanation of age checks. 18 minutes video.
A dangerous legislation is en route and is rushed so people won't have time to ogranise.
View quoted note →
When the "reboot" command sneaks into the wrong terminal 

There is this one version of our future where software does not exist. Literally - everything is done by LLMs.
You want to see a website? An LLM constructs a http query, fetches it and displays the content.
It seems like a nice solution. But not really.
The problem has already started - CENSORED LLMS.
For some reasom, the first commercial versions of LLMs were censored - and as a toy, it didn't really matter. But this trend still continues to this day. Tools should not be censored.
As knowledge databases, LLMs are already discrminating some people, including scientits. But it's still hard to call it a censorship at this point using a standard censorship definition.
The problem starts to appear when an LLM is supposed to act as an application/software and do stuff which such software should do for us. Because of the current heavy censorship of LLMs, such apps will deny their users basic rights, including the most important one: freedom of speech. You want to talk about nuclear weapons construction? Well, you're out of luck, the LLM acting as an app for an internet forum will deny the request. Any hardcore description, psychological help, medical imagery - everything DENIED.
The LLMs lawful alignment is problematic - the law is not static, which is a separate problem in itself, it's different all over the world, etc etc. So who decides what LLMs acting as apps will block or pass? My e-bike software which will be an LLM, may disable my bike if I ask it to help me ride to an abandoned quarry to smoke weed. I definetely do not agree for such future.
But there is hope. Maybe, but only maybe, currently abliterated and uncensored models, are not only toys to generate some fucked up text, but also a beginning of a resistance movement which at the core value will have a slogan: "I decide about my actions, not my LLMApps and software".
I think it's the humanity which should have power over the LLMs and AGI in general, not the other way around.
I would also like to point to extreme errors in law in the past, like prohibition, nazi germany laws and multiple other. Are we really sure that current laws are good? Maybe they are fucked up the same way as those old laws were, but we don't know it yet? Will our "software-composed-LLMs" try to enforce this bad law?
Why I'm mentioning this? Because someone says: BUT WE HAVE FREEDOM OF SPEECH WRITEN IN THE CONSTITUTION.
Well.. I have some bad news. As an example I will use the most dark western country which is dark exactly because of changes which I will describe in a moment.
So people think, that if there is a law, like the "free speech", that it's a true real law. It's not.
A good example in this area is the United Kingdom. There is or rather in their case, WAS a law, which people took for granted. A law which should protect people. But it turned out, when that particular law was problematic for the government - it was literally destroyed. And you may think it was something small, some non-important law. Here's the catch - this is one of THE MOST imortant laws in western countries. Which law I'm talking about? the "PROCTECTION FROM SELF-INCRMINATION". So yes - the UK broke that protection, by forcing the accused to deliver decryption keys, which is a direct translation to a "tell me where the notebook is or you are going to jail". And it doesn't matter if the notebook doesn't exist at all. Current courts are hesitant to use it, exactly because it breaks a lot of law rules and basic human rights, which may cause an extreme reparations being paid out in the future for those affected.
Should LLMs follow broken laws? Do we know which are bad and broken right now? Should people yield to what a "software-LLM" decides? Maybe we should! - maybe AGI should create rules for humans? But who will control AGI?
Are you ready to be controlled by transformer's layer weights?
He may be onto something here...
View quoted note →