I think the whole discussion about consciousness misses the simple point that LLMs might just work better if we treat them as if they were conscious.
Maybe it's a coincidence that the company doing this also tends to have the best models (and other factors certainly play a strong role). But I think it's plausible that focusing on "model welfare" actually makes models better at their tasks.
It's interesting to me that one can look back at the effects that decision has had on the US and say it "was 100% correct."
It's a bit like sitting in the burning ruins of Rome and contemplating that Nero was 100% correct to focus on his music. I mean, I'm glad he got to do what he loves, but maybe 100% is just a tiny bit of an overstatement.
It's more like if the law says the maximum sentence for theft is 10 years, a thief appeals his 20 year sentence, wins, and gets out early. He goes and robs somebody else so you say the court was wrong to let him win.
The court's job is to uphold the law. If you disagree with their interpretation, you can call them incorrect. If you have a problem with the consequences of the law, you have a problem with the legislature.
True, but in this case they made a determination about where the acceptable limits on a constitutional right fall which leaves quite a bit more room to disagree with them. It's not at all clear to me that spending money was intended by the framers to be unconditionally protected by the first amendment. We've even got the interstate commerce clause and IP law codified in the same document so how is that not an obvious inconsistency? IIUC SCOTUS based the distinction on the political nature of the activity but certainly that's not something spelled out in the original document.
Came here to say the same thing. Models used to rely heavily on world knowledge from their training data. They are now much better at tool use and deciding when to research a topic, rather than just answering from memory.
I wonder how much that extends to using LLMs for programming. I assume most knowledge of programming language syntax still comes from training data.
I find they generally do ok, but a few lines in an AGENTS.md or manual prompting to verify stuff against current docs/source, and check for current version of software helps a lot.
That won't be a problem for long. The native Europeans are swinging to the authoritarian right, the newly naturalized are very selectively for liberal democracy - aka the parts that directly benefit them, not the parts that make it liberal, and the liberal democrats are straight out refusing to have kids.
So in 50 years Tukey will border authoritarian regime either way.
It's interesting to me that people generally see Europe's regulations as interfering with entrepreneurship, while they see American decisions that give more power to multinational corporations at the expense of smaller companies as a normal part of a free market.
Starting a company is not the hard part of staring a company in any country. Many many companies do exist in the EU. It might be annoying but it's obviously hardly insurmountable.
Yes, I saw one just recently where the poster made obviously false statements about starting a company in Germany.
(Before you downvote this comment, please note that I have started multiple companies in the EU. I'm speaking from personal experience, not from something I've read on Hacker News.)
That's unfettered capitalism at work—"free" implies no bounds. Multinational corporations have never been more powerful now than at anytime in history.
And that hurts the free market. A free market should be transparent, have low barriers to entry, and not be dominated by monopolies or cartels. There's nothing free about American capitalism.
Your implied premise is that private companies cannot impede others' freedom, which isn't true. In unfettered capitalism, only the largest corporations have the power and money to be free.
> SWE-2 is free to use for users like yourself for the next month, and almost all usage should be supported via our CLI (https://docs.devin.ai/cli)
Thanks for pointing that out; I wouldn't have found out about it otherwise. I bought the 20$ subscription and have been using SWE-2 for a few days now, and it's actually good. I don't think it's Fable-class, but SWE-2 Max feels substantially better than K3 and on par with Opus 5 Max.
A hallucination in the context of LLMs is generally understood as an incorrect answer presented as factual. If you claim that "x can't hallucinate" in the context of LLMs, you're saying that x always gives accurate answers. It does not matter whether the answer is type safe. If its value is incorrect, it's a hallucination.
I only use chinese models for code reviews because you can actually tell them to take an adversarial stance and actively look for security issues without risking refusals. GLM-5.3 has been great for this, although it can be slow on larger PRs.
reply