Grimblewald

6 ms·

1.
by Grimblewald 3d ago
especially when emebeddings are enough to recover original signal if the model is built for it, which these no doubt are.
2.
by Grimblewald 3d ago
yes? im not sure what your point is, i sense you seem to disagree, but what youve said supports what ive said so im not sure how to engage.
3.
by Grimblewald 3d ago
to replacement, we're rapidly falling below. distributuon and related trends are extremely grim.
4.
by Grimblewald 4d ago
much harder to do in OP's case, matching flavour then referencing that specific companies name when asked how it know to flavour this way? thats astronomically low for randomly selected plausible tokens without some data prior, like th
5.
by Grimblewald 4d ago
because breeding intelligent creatures generally requires keeping them happy. When treated like a farm aniaml, many high cognition anaimals will refuse to reproduce, just look at how bad life has become for average humans, and predictably,
6.
by Grimblewald 5d ago
also why hyperscalers are paying folks to host gpu clusters in their homes. its to get at tgeir IP, it has nothing to do with space, cooling, etc.
7.
by Grimblewald 5d ago
llms right now work like pre cnn computer vision based on MLP's. By this i mean brute force of a model not really built for the task, and lacking a task specific inductive bias, being made work with unfathomable volumes of data and she
8.
by Grimblewald 5d ago
right, but that surive in isolation, heard immunity to collapse so to speak, but globe wide? pandemonium. we're back to fiefdoms and warring city-states in a generation.
9.
by Grimblewald 6d ago
plenty of ex-civs had techdoomers who were right. When tevh doomers have no salient points to make, its fair to not listen, but this time around i think something is different to historic doomerism. Trust is a fundamental thing, and i canno
10.
by Grimblewald 6d ago
I'm starting to have chinese characters bleed into claude as well. Perhaps a sign of the times. Understanable for a chinese first model but an english first (supposedly) model? wild stuff.
11.
by Grimblewald 7d ago
I really do think its tge case. Ever notice how juuust when you run out of usage its allllmost correct? then the handy helpful "buy more credits!" link pops up. Meanwhile local models are finishing the task without fucking around
12.
by Grimblewald 7d ago
people seem to miss tge point of this. The problem isn't about credit, its about portraying these models as more competant than they really are. It fuels idiotic statements like jensen huangs recent "agi achieved" statement,
13.
by Grimblewald 8d ago
no profit is very different from running at extreme loss, as an example. A company can be eternally stable with zero profit, while another very profitable vultur capitalist company will eventually run out of companies to gut for good lookin
14.
by Grimblewald 9d ago
my gripe is, isnt this filling the role pro was supposed to fill?
15.
by Grimblewald 9d ago
i'd disagree, hard. in no observable way is anything currently avaliable or demonstrated at a level one can call "agi" it's no secret that nvidia, under jesen's leadership, is in dire straits financially. This claim
16.
by Grimblewald 10d ago
public data and a well know situation innit?
17.
by Grimblewald 13d ago
i dunno about more expensive, but I do known it fixed several side projects opus 4.8 has been struggling to get working, using only their standard plan.
18.
by Grimblewald 13d ago
if no one used alternative stores, there'd be no reason to ban them, and yet google is banning them.
19.
by Grimblewald 13d ago
right but sideloading means installing outside the official channel. last i checked playstore was an option, not android official.
20.
by Grimblewald 17d ago
ok, if that's the case, my OS critique holds word for word if replacing with some libre flavoured "office" suite.
21.
by Grimblewald 24d ago
producing an os with a focus on utility to end user would be a start. Even just something that makes some form of effort to give users a return on investment on their hardware. Right now, you unluck massive headroom, battery life, and stabi
22.
by Grimblewald 26d ago
people have always, and will always, be stupid as fuck in any domain they dont make their life. We can handwring terminology all we like but it wont stop the mythologizing and misleading takes on the topic. Not one bit. So why not just use
23.
by Grimblewald 1mo ago
Personally I'm a fan of wsl, it has let me dip my toes and transition slowly into the environment, enough so over the last 4 years that this weekend im wiping windows from everything I control. Windows does nothing but bring me pain at
24.
by Grimblewald 1mo ago
think of it like the words "car" and "engine", one goes into the other and both have their impact on overall capacity, feel, performance etc. each in their own way. The llm is an engine, the harness is the rest of the ca
25.
by Grimblewald 1mo ago
where'd you read that? sounds like total bs but i could be wrong and would like to learn more.
26.
by Grimblewald 1mo ago
not really difficult. Genocide is a rather niche thing to like, and supporting it puts you in lockstep with those perpetrating.
27.
by Grimblewald 1mo ago
the back button hijack on mobile is a nice touch.
28.
by Grimblewald 1mo ago
This is a terrible idea. It sounds like the way they're going to do this is by looking at specific word choices, which is the standard say AI "detection" currently works. The problem is that if it is "undetectable"
29.
by Grimblewald 1mo ago
i tend to give ai a "voice" file, which it must use to write. I hope this doesnt spread to providers i do use, since if it does i mugbt just forgoe ai, i cannot stand ai prose.
30.
by Grimblewald 1mo ago
well, i's love to see what lead to mythos trying this. if instructed to do it, which it likely was, it doesnt prove much of anything.

More ›