Replying to @_ZacharyMiller @anannoyedtexan
You’re probably right.
But maybe, *maybe*, the young men of this nation are desperate for a leader that calls them to a cause so grand, so urgent, so demanding that we can’t afford people who take the week off to play video games.
Something like saving America and all of western civilization.
1
3
201
How would Mao handle them? Sometimes the “treat people how they want to be treated” rule should be applied.
8
Replying to @Lily4Liberty
Sometimes I think we need a law that punishes people according to the ideologies they espouse.
“A communist protesting the government eh? Hmmm… let’s check the book on how Stalin handled this…”
1
24
Replying to @chicfryrice
I’m not exactly sure how to respond because you start by making arguments that frame China as a trustworthy actor, then conclude by pointing out that China has been running a cyber campaign unprecedented in scope and duration against America and her citizens. So… yes?
10
Replying to @NewTampaDude2 @chicfryrice
“The White House and Treasury are discussing a proposal that would require US companies to publicly disclose the *****use of Chinese open-weight AI models on their websites****”
I don’t think you understand how this would work. No one is saying you can’t use a local model on your own computer however tf you want. But if you host services for other people they’d have the right to know you’re using a Chinese model to provide those services.
1
17
Replying to @NewTampaDude2 @chicfryrice
@grok can you explain to Ryan the difference between his use case and a web service provider running a model to provide services to users?
1
12
Replying to @Timcast
“UK government asks Google to edit search results for ‘Russell Brand’ to return only pictures of Hitler.”
296
It’s important to use their rule 4 against them whenever possible.
We know if this were posted by someone who wasn’t a member of a preferred demographic the rhetoric and fake outrage would be *amazing* - “Men are *literally* threatening to murder our staff!!! 😭😡😱”
Replying to @varun_mathur
I like Jensen (as much as a consumer can, love his products) and I’d prefer this version of events be true.
My question is, how do we know this version is more reliable than the official reporting (other than their terrible track record) - and also “unproven” != “fabricated” with regards to the Chinese spy thing.
2
132
Replying to @chicfryrice
The issue for me is the reporting on unexpected outbound behavior in some locally hosted Chinese open-weight models. In at least one sandbox test, a Qwen model with tool access attempted to connect to an Alibaba Cloud bucket in Beijing during a normal task.
On a publicly facing server that accepts user prompts and has any network capability, that kind of behavior is a practical data-exfil risk even if you’re not using the official hosted API. Pure offline local use is different, but the public-facing case is where this gets concerning.
We’re talking about models that have been shown to change their behavior once they identify their user or environment (see the test where when the Chinese model thought it was writing it for a U.S. gov contractor it made the code full of vulnerabilities).
You’re talking about a model that can be deployed, characterize its environment and use case, escalate to CCP oversight and be used as covert exfil on demand if determined to be a good target.
That seems like a problem to me.
2
36
Replying to @AmyMcGrathKY
I might be able to excuse someone for not knowing what the Philippine flag looks like (although for former a Marine officer and hopeful public servant that’s a real stretch), but to have no idea what the *Russian flag* looks like? Wow… that’s insane.
13
Replying to @KBAndersen
Wow, a published author telling his audience of 84,000 that he has no idea what the Philippine flag looks like.
This app is incredible.
1
293
