firejake308

Conclusion: "The binding constraint appears to be engagement: realizing the promise of AI tutoring will require getting students to use it, not just giving them access."

Yes, but the generalists are not routinely improving across all domains. The large labs are really focusing on agentic use, so I imagine that creative writing has deteriorated considering how distinctive Claude's writing style has become. Or I recently had an image-parsing task, and I was excited to try Qwen because I heard it had gotten a lot better at agentic tasks, but it failed my image-parsing benchmark.

> The operational signal was always relative preference. The scalar merely hid it.

Is this another Claude-ism? "X was always Y. The Z merely hid it." Or am I overcalling it?

Counterargument: this works for quick prototyping, but for any serious business, you will eventually develop a benchmark/eval to track how well the general model is working, and once you have that dataset, you might as well train a specific model

> Maybe the goal is to force the U.S. Census to publish statistics that actually enable re-identification, to help with future gerrymandering efforts?

In case you were wondering why the government would do this, yes, that's exactly why.

> The new SECURE proposal includes language ... [that does] not allow people to sue companies for violating their privacy rights....

In case you wanted to know what this was about

Right, this is what I remember seeing. Overall, inflation didn't increase much last year even after the tariffs because a lot of companies had stockpiled inventory and were able to postpone the price increase. Seems like that bet paid off, and now they should ideally be able to get their money back without increasing prices.

Obviously there were some companies who did raise prices because of tariffs, but I'm saying that on average, they must not have since inflation didn't go up by 15%

I mean, we need all the electricity we can get to run all of these datacenters, right? So I think this might be one of those things that the Republicans quietly allow to continue so that the corporate interests can maximize electricity production

The LLMs are not uncorrelated, though, they're all trained on the same dataset (the Internet) and subject to most of the same biases

> “DeepSeek is more humane,” my mother told me in May. “Doctors are more like machines.”

Now there's a quote

> As of Saturday, the U.S.-based Human Rights Activists News Agency said it had confirmed 5,459 deaths and is investigating 17,031 more.

The 30,000 number comes from the Ministry of Health. It seems the UN number also aligns with the new 30,000 number. This is much worse than the 3,000 that was reported earlier. But it also seems like the crackdown is over now, and we're still just counting deaths from Jan 8 and 9.

I compare this to the recent protests in Bangladesh, where Sheikh Hasina ordered the military to shoot the protesters and the military refused. The difference between these two countries is proof that people do have the ability to disobey orders from authoritarian leaders, and that decision can have a huge impact.

If hallucinated citations are making it to top conferences, do we just have to accept that no one is willing to do the work to ensure that our research is grounded in reality? Perhaps the volunteer peer review system is broken and we need people who are paid to carefully check the citations

If hallucinated citations are making it to a top conference, Is there anything we can do to stop this rampant abuse of AI? Or do we just have to accept that research is no longer going to be grounded in reality?