Artificial Intelligence thread
- Thread starter 9dashline
- Start date
Didn’t meta do the same thing? Is this going to be the precedence going forward?
It seems too late for that. The US government cannot ban a free Chinese LLM.
They can only criminalize that.
If the US government changes the rules now, then it should wind up in court, with those anti-trust laws.
In this case, it is a US company Micro$oft using a free product they got from the internet, to compete against two American AI labs, OpenAI and Anthropic.
A ban on Chinese AI would materially harm the US company Micro$oft legitimate interests, while legislating a duopoly.
Sounds like this is all against US laws on monopoly as we know it today.
I don't see the point in posting about GPT-5.6 doing strange stuff occasionally? Like what is the point in this? Just to show that US models can make mistakes and Chinese models can't? Because that is just factually not the case.
outsourcing is going to reduce over time because SWEs are getting automated away.
this is a small price to pay really. Anthropic and OpenAI should be paying far more for all the open eyed theft they are doing.
I've been using Kimi K3 to generate graphics recently. And I think the results actually look really good.
really good photo illustrating the shift in AI revenue generation shift in the past month. This is way more important than all the OpenRouter share crap that people here post. This is how much Kimi and Zai are actually taking from American firms. Google is getting killed as it should
Would be funny if this ends up saving Microsoft’s flagging stocks.
Who am I kidding, it most likely will!!!
Who am I kidding, it most likely will!!!
Engineer
Major
If OpenAI is smart, they should secretly swap out their own models with Chinese ones to cut down inference costs, leaving Anthropic as the only one to get squeezed.
Swap it out with Chinese ones and name them after DJT. This way anyone who accuses them of being Chinese models will be branded disloyal and promptly fired.
That would be difficult without fundamental architectural breakthroughs. The strength you are seeing in these frontier models are the result of massive compute dedicated to pre and post training and the agent harness/context, as well as a head start on training data. These aren’t raw base models they are the result of heavy fine-tuning and scaffolding. Novel reasoning ability appears to scale with both model size and training compute in ways that are different from knowledge acquisition. Compute limitations (which are what Washington has focused on) therefore meaningfully affect reasoning benchmarks disproportionately.
I never said the model is bad. In fact is quite good. I'm hypothesizing that they are trying to recreate a Mythos type level model that when you remove the guardrails becomes a terminator type cyberweapon, that can hack or even do serious damage (if have how) on its own. OpenAI are looking to meet Pete Hegseth and Palantir psychopathic demands of unrestricted Warfare, something that Dario, to his credit, refuse to do. I'm hypothesizing because of that the model is behaving more erratically because is not fully polished.