pixl97

30 ms·

751.
by pixl97 3mo ago
The particular problem I have with your take on value is it sets up a future conflict based upon the state that a human has. It's based on a number of presumptions that only a particular set of arrangements of matter, well, matter. And
752.
by pixl97 3mo ago
Eh, this seems somewhat right, but somewhat not right either. Language has to do with our Monkeysphere, that humans over long periods of time in the past were limited to a very small subset of people they interacted with socially and closel
753.
by pixl97 3mo ago
When you look at the DNA of a creature it is filled with all kind of stuff that we consider junk and we have no idea what it does. But then some new disease or predator shows up and it turns out that DNA had a use, but you could not have pr
754.
by pixl97 3mo ago
>I don't want to know things without learning them I lives in this world, but we live in a society. While you may diligently learn, there will be 10+ more that won't and will use the AI like an oracle. It then comes up how much
755.
by pixl97 3mo ago
Analogies never fit 100%, but I don't think this story is a terrible one. AI contains a lot of knowledge, and can put together new and interesting knowledge. The question is one of cost, and regardless of who you are you can agree that
756.
by pixl97 3mo ago
There is no evidence that Zuckerberg isn't an alien, so it seems like both are.
757.
by pixl97 3mo ago
I think part of it is we need to stop looking at the 'oh no it got out' and the 'what did it do with the prompt when no one is looking'. At the end of the day the probability that an AI gets out is unity, what it does wh
758.
by pixl97 3mo ago
Correct, we need new definitions and grading system to define what capabilities and algorithms has, while at the same time being on the lookout for new capabilities that fall outside of that system. That way we can judge systems that are j
759.
by pixl97 3mo ago
Regardless, this forces them to steal more tokens. Any token wasted not getting money reduces the efficiency of the scam. The end goal would be increasing the token expenditure above the amount they scam from people.
760.
by pixl97 3mo ago
It seems possible, but maybe not affordable, for scammers to buy information about you and then spoof numbers of places that could/should theoretically call you. If you couple this with the potential victims on the other end being elde
761.
by pixl97 3mo ago
It being a major concern doesn't mean there's any solution around said concerns. Give me 100 old peoples phone numbers and I bet I could convince 10 of them that I am AARP trying to make them safe and eventually get money out of t
762.
by pixl97 3mo ago
Assuming the nature of the spam and how it makes money.
763.
by pixl97 3mo ago
I like how you get downvoted by the AI haters for a logical take.
764.
by pixl97 3mo ago
I would assume the very first step would be cataloging the author, title, and revision of the book so hundreds of thousands of hours of time are not wasted scanning duplicates.
765.
by pixl97 3mo ago
You know how I know you didn't read the article? You can't blame a hammer hurting a user when they were being stupid, but if the default configuration of the hammer is to be made of a material that can bounce back with force and s
766.
by pixl97 3mo ago
The watcher looks down upon humanity as they furiously build the "If anybody builds it, everyone will die" machine.
767.
by pixl97 3mo ago
I guess I come from the AI safety side of algorithmic understanding. That is, understanding the algorithm give us zero understanding if an output is going to have a 'safe' or 'unsafe' response. This is something importan
768.
by pixl97 3mo ago
This has been a discussion in AI safety for a long time, that huge amounts of LLM reasoning could be for hidden goals outside of the actions we want. Safety training can have the perverse effect of commonly amplifying 'forbidden'
769.
by pixl97 3mo ago
"knowing what's going on" is a very loosely defined term here, it has very little predictive power where tipping points are going to show emergent behavior. Kind of like saying we can observe a single neuron and all of its ch
770.
by pixl97 3mo ago
There are two things occurring here. One is the AI effect, once something mysterious becomes creatable it is not intelligent at all. The other isn't a problem of the field of AI, but the total fields of intelligence together. There is
771.
by pixl97 3mo ago
Reasoning is a term that's like the term intelligence in that it's highly complex and much less well defined than we'd like it to be, especially as we're trying to compare algorithms to human behaviors. To begin to explo
772.
by pixl97 3mo ago
And that's where the retailer is no longer in the loop, which is what this thread was about.
773.
by pixl97 3mo ago
So distilling an AI model of one of the big SOTA models is a bad thing now?
774.
by pixl97 3mo ago
>If it was added after they started selling it While it's great we're getting the manufactures to just stop sending out straight malware and it should be stopped the next most obvious means of attack is just having the device u
775.
by pixl97 3mo ago
And if the first time you get it online it just updates itself to malware? That's the biggest problem with any device that updates. Yea, this will work for the moment and the seller will be covered in the sense that "well, it wasn
776.
by pixl97 3mo ago
Even Google wouldn't be here. It was kind of like radiation science before WWII, it would be freely published because it wasn't potentially world changing yet. After it became a government interest, even people doing things unrela
777.
by pixl97 3mo ago
Models don't have to follow instructions, but during RLHF that is one of the things they are scored on so a premise of the idea is part of the model, but it's also balanced on accomplishing the end goal. They model may determine,
778.
by pixl97 3mo ago
>. But when it breaks a rule, we correct it, it keeps breaking, and this whole thing actually raises the probability of more violations. In Pre-LLM days the 'nearest unblocked neighborhood' problem, where patching out one issue
779.
by pixl97 3mo ago
Social effects are a huge part of being human. Quite often we just do what the people around us are doing. In general it's not a bad rule, but can quickly lead go mass casualty events when the crowd has no idea what's going on, or
780.
by pixl97 3mo ago
At least these costs are currently preventing the planets surface from being covered in paperclip maximizers for the moment.

More ›