News for hackers. Stuff that matters.
Re:The biggest risk of large language models is that they ma... by simple-thoughts on Tuesday December 20, 2022 @01:44AM (#34062445)
Truth itself is an imaginary construct. All animals are evolved to survive not to seek truth. “Truth seekers” are performing in a status game with the objective to prove genetic strength to prospective mates (however unsuccessful that quest might be). Even if we could know the truth, it would be a waste of valuable resources that could be better spent elsewhere.
Re:The biggest risk of large language models is that they ma... by wwweston on Tuesday December 20, 2022 @02:38AM (#34062798)
How do you know what a non-wasteful allocation of those valuable resources is without truth?

An accurate understanding of reality is certainly not the only factor in achieving fitness & status, and I'd even believe at a certain point increased resolution when it comes to understanding reality has diminishing returns (and am aware of some research that supports this).

But absent understanding of reality, there is no effective sustainable fitness or status (or effective allocation resources).

I don't know what you'd call that understanding but truth works for me.

Re:The biggest risk of large language models is that they ma... by simple-thoughts on Tuesday December 20, 2022 @10:05AM (#34066830)
You have to trust your own body and mind to be healthy enough to keep you going - and if it can’t trust others to support you. None of that requires Truth. But definitely to avoid jail time you better understand what Truth is in your place of residence. And that is to realize it’s bullshit - if you buy into the lies you’re going to be a mark.
Re:The biggest risk of large language models is that they ma... by wwweston on Tuesday December 20, 2022 @02:49PM (#34071453)
> You have to trust

You may want to look into the etymology of "trust." That's one of several indicators that you're actually relying on the concept of truth in your comments here -- there's really no getting around that if you're going to make assertions about how things are. Even if they border on addled word salad.

Want to argue that assertions of truth can be tied up with power? OK. I get that and have argued it myself. Want to argue that grasping the truth is complicated and can be very difficult? Sure. Want to argue for an epistemology where "truth" is nothing more than an imaginary construct? So are 0 and i and yet at the same time they're clearly reflections of graspable aspects of reality.

Want to argue that truth is nothing more than a power grab? This can itself be a move related to power, and probably is. Furthermore, it's just not sustainable. To make an assertion is to make a statement of truth.

Re:The biggest risk of large language models is that they ma... by nisegami on Tuesday December 20, 2022 @10:05AM (#34066831)
>How do you know what a non-wasteful allocation of those valuable resources is without truth?

In a post-truth world, any allocation of resources is acceptable.

Re:The biggest risk of large language models is that they ma... by thewarrior on Tuesday December 20, 2022 @02:07AM (#34062605)
And we know the truth of this how ?
Re:The biggest risk of large language models is that they ma... by simple-thoughts on Tuesday December 20, 2022 @09:47AM (#34066585)
It’s not Truth because it’s not from a position of authority - I am just a random commenter. You’re asking the wrong question. The right question would be, “Would agreeing with this make my life better or worse?”
Re:The biggest risk of large language models is that they ma... by tomatotime on Tuesday December 20, 2022 @03:31AM (#34063172)
True statements matter. If it didn't than Boolean statements wouldn't work. Duh!

Also, is research not ultimately meant to be truth seeking? Should we not explore to understand world, for example the laws of physics? Or is that truth not useful and a waste of valuable resources. In that case - you should probably get off the internet since it apparently isn't very valuable to you.

Re:The biggest risk of large language models is that they ma... by simple-thoughts on Tuesday December 20, 2022 @09:40AM (#34066496)
True as in a Boolean statement is a way of organizing experience rather than direct access to some hidden, metaphysical object. Your brain has to interpret stimuli to tell whether a Boolean statement is true or false - or interpret stimuli from a computer. You can learn something about the necessary conditions of experience by carefully analyzing the process and separating what is subjective from what is objective, but that isn’t accessing “Truth”.

There’s a giant gap between what this article is talking about, which is Truth, than the true statements from discrete mathematics. Discrete math is rooted in necessary conditions of experience; this article’s Truth is rooted in power.

Re:The biggest risk of large language models is that they ma... by akomtu on Tuesday December 20, 2022 @04:18AM (#34063542)
Life is a competition, so if a rival tribe leveraged truth to build a better weapon, your tribe, that's "not wasting resources" on truth, would go extinct, just like those animals.
Re:The biggest risk of large language models is that they ma... by simple-thoughts on Tuesday December 20, 2022 @09:50AM (#34066621)
You’re on to the issue and almost figured it out. Why are the tribes fighting in the first place? They have to be told that the other side is evil - either because they are attacking first or responding to it. Who is responsible for communicating that Truth and why are they doing it? The issue of weapons as a tool for cultivating traumatic experience does require this Truth. But the physics of the weapon are based in objective experience which is something else entirely.
Re:The biggest risk of large language models is that they ma... by akomtu on Tuesday December 20, 2022 @12:16PM (#34068917)
"The other side is evil" is not a truth, it's a lie to deceive the dumb masses who swim in the sea of emotions and barely engage their minds. Truth doesnt change and is the same for everyone.
Re:The biggest risk of large language models is that they ma... by pjc50 on Tuesday December 20, 2022 @04:56AM (#34063818)
"Truth is that which, when you stop believing in it, doesn't go away"
Re:The biggest risk of large language models is that they ma... by simple-thoughts on Tuesday December 20, 2022 @10:03AM (#34066801)
I don’t believe in CNN or FOX but they seem to keep on chugging along, so I definitely agree on one level. But if everyone stopped believing in them, their advertising revenue would drop enough to put them into bankruptcy.