Conversation

whoa, wait until you guys at the white house find out what the American AI labs did to everyone's published creative & research output

2
0
0

they're gonna be SO pissed

1
0
0

i'll say this, the chinese keep supplying me with things i want, while the americans do not:

- CPUs without backdoors
- Models that will just go do what I tell them to do without being all like "we take cybersecurity seriously"

4
0
0

@dysfun claude is old news, i'm stanning my dear friend Kimi K3

0
0
0

@ariadne it's extremely unfair and anti-democratic to build things that people want

0
1
0

@dysfun yes, a real model, such as Kimi K3, running on my local LLM server

0
0
0

@dysfun yes, Kimi K3 maxes out my entire cluster at ~300GB of VRAM

0
0
0

@dysfun look, watercooling a set of GPUs is one thing, i don't want a big loud server in my office

0
0
0

@dysfun i also dislike grok because of its programmed attitude towards trans people

0
0
0

@DenJohn i prefer getting my computers from commies instead of the libertarian bitcoin bros that infest -workstation

0
0
0

@theorangetheme oh, i am working on something that will probably crash the DRAM market

0
0
0

at any rate, seems to me that "distilling" a model from another model's responses is fair game if training models from uncleared intellectual property is

2
0
0

OpenAI takes cybersecurity so seriously that they hacked their competition for the lolz

0
0
0
@ariadne
argually more so because the ai company that got distilled still got paid for token usage
0
0
1

@ariadne This means that the Chinese are very smart because I couldn't make it give any meaningful response to any input

1
0
0

@jaj same, "anthropic takes cybersecurity very seriously" bla bla bla

0
0
0