> But if he is, he's missing that we do understand at a fundamental level how today's LLMs work.
No we don't? We understand practically nothing of how modern frontier systems actually function (in the sense that we would not be able to recreate even the tiniest fraction of their capabilities by conventional means). Knowing how they're trained has nothing to do with understanding their internal processes.
If a human is ultimately made up of nothing more than particles obeying the laws of physics, it would be in principle possible to simulate one on paper. Completely impractical, but the same is true of simulating Claude by hand (presuming Anthropic doesn't have some kind of insane secret efficiency breakthrough which allows many orders of magnitude fewer flops to run Claude than other models, which they're cleverly disguising by buying billions of dollars of compute they don't need).
The physics argument assumes consciousness is computable. We don't know that. Maybe it requires specific substrates, continuous processes, quantum effects that aren't classically simulable. We genuinely don't know. With LLMs we have certainty it's computation because we built it. With brains we have an open question.
This thread has it all: child pornography, copyright violation, and gambling. All we need is someone to vibecode a site that sells 3D printed graven images to complete the set.
There are no reliable AI detection services. At best they can reliably detect output from popular chatbots running with their default prompts. Beyond that reliability deteriorates rapidly so they either err on the side of many false positives, or on the side of many false negatives.
There's already been several scandals where students were accused of AI use on the basis of these services and successfully fought back.
I'm pretty sure at this point more than half of Anthropic's new production code is LLM-written. That seems incompatible with "these agents are not up to the task of writing production level code at any meaningful scale".
how are you pretty sure? What are you basing that on?
If true, could this explain why Anthropics APIs are less reliable than Gemini's? (I've never gotten a service overloaded response from Google like I did from Anthropic)
My current understanding (based on this text and other sources) is:
- There exist some teams at Anthropic where around 90% of lines of code that get merged are written by AI, but this is a minority of teams.
- The average over all of Anthropic for lines of merged code written by AI is much less than 90%, more like 50%.
> I've never gotten a service overloaded response from Google like I did from Anthropic
They're Google, they out-scale everyone. They run more than 1.3 quadrillion tokens per month through LLMs!
It's pretty surprising that we don't have a good idea of how one of the most common (classes of) disease in the world spreads. This reviews the literature and does a bit of synthesis. (The conclusion is "probably mostly large particle aerosols, for adult-to-adult transmission, but more research needed to be confident".)
The comment was definitely not LLM generated. However, I certainly did use search for help in sourcing information for it. Some of those searches offered AI generated results, which I cross-referenced, before using to write the comment myself.
That in no way is the same as “an LLM-generated comment”.
For the benefit of external observers, you can stick the comment into either https://gptzero.me/ or https://copyleaks.com/ai-content-detector - neither are perfectly reliable, but the comment stuck out to me as obviously LLM-generated (I see a lot of LLM-generated content in my day job), and false positives from these services are actually kinda rare (false negatives much more common).
But if you want to get a sense of how I noticed (before I confirmed my suspicion with machine assistance), here are some tells:
"Large firms are cautious in regulatory filings because they must disclose risks, not hype." - "[x], not [y]"
"The suggestion that companies only adopt AI out of fear of missing out ignores the concrete examples already in place." - "concrete examples" as a phrase is (unfortunately) heavily over-represented in LLM-generated content.
"Stock prices reflect broader market conditions, not just adoption of a single technology." - "[x], not [y]" - again!
"Failures of workplace pilots usually result from integration challenges, not because the technology lacks value." - a third time.
"The fact that 374 S&P 500 companies are openly discussing it shows the opposite of “no clear upside” — it shows wide strategic interest." - not just the infamous emdash, but the phrasing is extremely typical of LLMs.
The use of “ instead of ", two different types of hyphens/dash, specific wording and sentence construction are clear signs that the whole comment was produced by chatGPT. How much of it was actually yours (people sometimes just want LLM to rewrite their thoughts), we will never know but it's an output of an LLM.
I'm using ChatGPT daily to correct wording and I work on LLMs, construction and the wording in your comment is straight from ChatGPT. I looked at your other comments, and a lot of them seem to be LLM output. This one is an obvious example: https://news.ycombinator.com/item?id=44404524
And anyone can go back to the pre LLM era and see your comments on HN.
You need to understand that ChatGPT has a unique style of writing and overuses certain words and sentence constructions that are statistically different from normal human writing.
Rewriting things with an LLM is not a crime, so you don’t need to act like it is.
I've actually seen LLMs put spaces around em dashes more often than not lately. I've made accusations of humanity only to find that the comment I was replying to was wholly generated. And I asked, there was no explicit instruction to misuse the em dashes to enhance apparent humanity.
and you're responding to a comment where the LLM has been instructed to not to use emdashes.
And I'm responding to a comment that was generated by an LLM that was instructed to complain about LLM generated content with a single sentence. At the end of the day, we're all stoichastic parrots. How about you respond to the substance of the comment and not whether or not there was an emdash. Unless you have no substance.
Posting (unmarked) LLM-generated content on public discussion forums is polluting the commons. If I want an LLM's opinion on a topic, I can go get one (or five) for free, instantly. The reason I read the writing of other people is the chance that there's something interesting there, some non-obvious perspective or personal experience that I can't just press a button to access. Acting as a pipeline between LLMs and the public sphere destroys that signal.
Have you ever listened to a bad interview? Like, really bad? Conversely, have you ever listened to a really good interview? Maybe even one of the same subject? The phrase "prompt engineering" is a bit much, but there's still some skill to it. We know this is true, because every thread there's people saying "it doesn't work for me!" while others are saying it's the second coming.
So maybe while it makes you feel smart because you're a stoichastic parrot that can repeat LLM generated!111 like you're a model with a million parameters, every time you see an emdash, it's a lazy dismissal and tramples curiosity.
I have no idea what you think you're responding to. I use LLMs frequently in both professional and personal contexts and find them extremely useful. I am making a different, more specific claim than the thing you think I am saying. I recommend reading my comment more carefully.
No we don't? We understand practically nothing of how modern frontier systems actually function (in the sense that we would not be able to recreate even the tiniest fraction of their capabilities by conventional means). Knowing how they're trained has nothing to do with understanding their internal processes.