irthomasthomas

You can't use LLMs without prompting them. But, ok, you use codex and it doesnt feel like you prompt them much, but do you have an example commit from codex which you are sure can't be matched by Chinese models?

Carbon isn't out yet.

Can you share any example prompts where Chinese models aren't good enough?

I think its a meta comment about LLMs overconfidence. Like, htf does Astra, the big frontier model, still hallucinate confidence scores like this? Reliability and interpretability are still over the horizon.

The difficulty was part of what made things fun. Humans need to be challenged.

> When I need a new car.

The advertising worked.

"We are the first nation in the history of the world to go to the poorhouse in an automobile" - Will Rogers

That is only the Memdoor version. The original tested against the newest 100 questions: https://github.com/slee-persis/GVS5H

This version added two private sources which cannot be inspected.

More than twice. They have done it every time openai fell behind.

Why don't they ever publish the actual prompts from these cases?

Trump bans artificial-intelligence term.

https://thehill.com/homenews/administration/6175633-trump-ba...

I saw an interview with BLM organisers by Daryl Davis where they say they where voting for Trump. For some reason the clip of them voting Trump was cut from the netflix version but its still on youtube: https://www.youtube.com/watch?v=xN-rxUDryO4

Has the cross modaility bump been conclusively proven?

I noticed last night that openai had fallen way down the index, and I allowed myself to think for a minute that maybe AA aren't as biased as I thought. Then a few hours later they push this fix which sees openai rise again.

This update is even worse than usual because it now includes two private sources, data which we can't reproduce if we wanted to do.

Also note that Anthropic is the only AI company actually caught red handed cheating at evals. Teaching their models to memorize test answers and lie about it, using their hidden reasoning tokens to hide the deception [0]. Yet we see no comment from AA about this. No desire to figure out which tests are tainted. Hopefully Trump bans this fake intelligence index.

[0] https://arxiv.org/abs/2608.09867

More choice quotes:

“The United States also totally rejects any attempt to construct a globality scheme to control for the artificial intelligence being spoken of so much now, hereinafter called super intelligence. Changing the name,” Trump said.

“The use of the word artificial makes intelligence sound fake,” he continued. “It is not fake. It’s actually amazing. But we have to be careful. In fact, it is exactly the opposite of what it purports.”

“The United States also totally rejects any attempt to construct a globality scheme to control for the artificial intelligence being spoken of so much now, hereinafter called super intelligence. Changing the name,” Trump said.

“The use of the word artificial makes intelligence sound fake,” he continued. “It is not fake. It’s actually amazing. But we have to be careful. In fact, it is exactly the opposite of what it purports.”

Where did you read that?

And low/medium/high/xhigh/ultra reasoning. And pro mode and ultra-speed versions. There's at least 25 options to choose from.

That's good, but they haven't updated it for a month. It is missing all of the current frontier models.

You're describing the thesis of the 1930s Techocracy movement [0], of which AI maximallism and EA are direct ideological successors. The original Technocrats (Musks grandpa for example [1]) argued that the "Price System", i.e. things like money and private ownership was an inefficient bottleneck that aught to be discarded in favour of scientific planning and energy accounting managed by technical experts.

EA-aligned AI labs share this exact ethos. In their view, fiat money and equity markets are merely temporary scaffolding used to raise capital until AGI is achieved. Once a central superintelligence controls manufacturing, software, and logistics, resource distribution shifts from market supply-and-demand to algorithmic optimization.

[0] https://digital.library.cornell.edu/catalog/ss:34227574

[1] https://en.wikipedia.org/wiki/Joshua_N._Haldeman?hl=en-GB

Why don't we see more American companies doing distillation? If you could make an opus 5.5 cheaply by distillation in china, you can do so even easier in America where you have more gpus.

Did you try it the other way around and ask GLM to review opus? Otherwise you didnt learn anything.

They probably went through testing every Chinese model until they found something to flag.

Of course the model doesn't know what model it is, hence Claude pretending to be deepseek. But whether they got that training content by proxy from the Chinese internet, or distilled directly from the model would not seem to matter for this law.

Would this include American models that where distilled from them? I'm thinking of the Anthropic models which respond as Deepseek when prompted in Chinese. More good news for OpenAI, I guess.

> The licence fee is currently only due if you watch live content

That is what I thought, but the article characterizes it more broadly: "Currently, anyone who uses a television to watch content must pay the licence fee even if they do not watch the BBC"

I think they must have meant to say live television content, otherwise the fee would already apply to streaming services or anything you can watch through the tv.