SkyBelow

9 ms·

61.
by SkyBelow 5mo ago
What's the chance that it is market motivated? That the companies most likely to succeed are those willing to break the rules (this isn't to say that breaking the rules makes one likely to succeed, you have to break the right rul
62.
by SkyBelow 6mo ago
Technically LLMs can be ran in deterministic mode as well, but I don't think that is enough. Even a deterministic LLM is too chaotic, small changes in prompts or the otherwise general context can result in vastly different outputs. T
63.
by SkyBelow 6mo ago
As much as I use AI, even for coding, I really do not like the argument. They are too chaotic to be compilers. The descent from prompt to code has far too many branches, and even small requests begin to build up bad patterns. There is som
64.
by SkyBelow 6mo ago
Even with all training data provided, won't it still be a black box? Unless one trains it exactly the same, in the exact same order for each piece of data, potentially requiring the exact same hardware with specific optimizations disa
65.
by SkyBelow 6mo ago
One issue with factual reporting is what facts are getting reported, given that public attention is a very limited resource. People consistently extrapolate from data without knowing if that data is good or bad. So if I show you news with
66.
by SkyBelow 6mo ago
Wait, I thought we were onto racoons on e-scooters to avoid (some of) the issues with Goodhart's Law coming into play.
67.
by SkyBelow 6mo ago
It seems most takes on this are that ends either always or never justify the means, but rarely is their discussion on the option that they can and developing a system of when they do and don't. At least in the general public discourse
68.
by SkyBelow 6mo ago
Can't one get randomness and determinism at the same time? Randomly generate the data, but do so when building the test, not when running the test. This way something that fails will consistently fail, but you also have better chance
69.
by SkyBelow 6mo ago
Could it be that stars were a good proxy, but as people realized such, they started being gamed, resulting in them becoming a bad proxy? Goodhart's Law would seem to always be in play for any proxy, because once it is recognized as a
70.
by SkyBelow 6mo ago
So, how would it apply to web searches? If a lawyer searches something for a person's case, is it protected? If a person searches something for their own case, does it have a similar level of protection? Seems AI chats would need to
71.
by SkyBelow 6mo ago
It needs to be something stronger than just deterministic. With the right settings, a LLM is deterministic. But even then, small variations in input can cause very unforeseen changes in output, sometimes drastic, sometimes minor. Knowing
72.
by SkyBelow 6mo ago
You can try to set up a NN where some of the neurons are either only activated off of 'safe' input (directly or indirectly from other 'safe' neurons), but as some point the information from them will have to flow over in
73.
by SkyBelow 6mo ago
Scale is very different, but I wonder if human trust isn't the real issue. We trust technology too much as a group. We expect perfection, but we also assume perfection. This might be because the machines output confident sounding an
74.
by SkyBelow 6mo ago
>Would it be a mistake to use Desmos in a math classroom Maybe. Back in the day I had classes where we had to learn the rough shape of a number of basic functions, which built intuition that helped. This involved drawing a lot of them
75.
by SkyBelow 7mo ago
NFTs were fueled by two different drives. One interested in the technology and if it could do something new and interesting, and another seeing it as an area of speculation (be that fueled by get rich quick and cash out or thinking it is a
76.
by SkyBelow 7mo ago
>this codebase already has this logic At first glance this looks like it might be the halting problem in disguise (instead of the general function of the logic, just ask if they both have logic that halts or doesn't halt). I think
77.
by SkyBelow 7mo ago
For some of these, we fully disallow company to child transactions or interactions. Wouldn't applying the same logic require the adult to fetch any internet content and give it to the child on a case by case basis? In this case, it is
78.
by SkyBelow 7mo ago
My main issue with this argument is that we never applied it to any other age gate created in the past. Why? Maybe part of it was control, but also because we know parents will fail their kids, and there were cases bad enough in the past
79.
by SkyBelow 7mo ago
That really depends upon the the reputation. If the police were to arrest a lot of 'very dangerous people', and all you see on the new is that they put the 'very dangerous people' in prison, then that isn't going t
80.
by SkyBelow 7mo ago
>so what do they do, force you to quit and then sign up again? That adds a lot of friction. Maybe that's a good thing? Imagine if changing the T&C required cancelling everyone's account and then letting people sign up with
81.
by SkyBelow 7mo ago
It gets worse x2: the executor of the estate having signed up for Disney+ means the estate of the deceased loses the right to sue, despite the deceased having never signed up. Like a client being bound by all unrelated legal agreements the
82.
by SkyBelow 7mo ago
>if the end result is objectively the same The issue is how this will be handled in law. Can the law define this in a way that is not overly strict or overly permissive? The current attempts is effectively the law doing this, but with
83.
by SkyBelow 7mo ago
In some sort of weird sense, it makes me appreciate the 'free armor trimming', 'alt F4 helps block attacks in pvp', and similar people in RuneScape. It gave young me a very low stakes environment to learn about scams, l
84.
by SkyBelow 8mo ago
>not status orientated Should we view those who chase status as a bad thing, or look to those who assign status that is then chased? If the average person cares more about who won last night's big game than some work done to improv
85.
by SkyBelow 8mo ago
Did it ever leave? I thought it was still preferred in certain medical procedures. I think something about after reattaching parts of the body it was one of the best options to ensure blood flow in the area, but it has been years since I
86.
by SkyBelow 8mo ago
If we allow FTL information exchange, don't we run into the possibility that the FTL accessible universe is infinite, so unique IDs are fundamnetally not possible? Physics doesn't really do much with this because the observable u
87.
by SkyBelow 8mo ago
Isn't this also the tactic used by someone who has been falsely accused? If one is innocent, should they not deny it or accuse anyone claiming it was them of being incorrect? Are they not a victim? I don't know, it feels a bit l
88.
by SkyBelow 8mo ago
>"One regulation was kinda bad, so we should never regulate anything again." Please don't misrepresent what someone says. That does not lead to constructive dialog. I gave a question challenging a specific way to regulate
89.
by SkyBelow 8mo ago
Because you have to be able to prove it wasn't AI when the law is tested, and keeping records and proof you didn't use AI is going to be really difficult, if at all possible. For little people having fun, unless you poke the wron
90.
by SkyBelow 8mo ago
It is worse, even less than useless. With the California case, there is very little go gain by lying and not putting a sticker on items that should have one. With AI generated content, as the models get to the point we can't tell any

More ›