scratchyone

This is super nice, thank you for including an AI disclosure. I would probably normally avoid something like this bc I would be considered by how much code oversight there is. Very nice to know that it's overseen properly by a human. Installed it and it's quite nice!!! great work :)

It would also be nice to be able to hide the checkbox it adds to the homepage. also disabling show focus box doesnt actually seem to work?

is there anything interesting in the unstable branch??

also lmfao bikeshedding a custom language is EXACTLY what i would expect from the dev of that kinda nerd game. feels like a good sign tbh

tbh they really didn't, tinygrad's was clearly a joke response. they were not providing a real uptime target.

I mean more fundamentally, if they have access to even more advanced models than all of us and have this much downtime, does that imply that their models are possibly not so great at software dev?

But yes you're definitely right, it's perhaps more ironic than contradictory.

Agreed. Having some level of human input makes a submission at least meaningful. If the entire repo and all text is generated by an LLM, does it really matter if the human is the one posting the link? It's functionally indistinguishable from automated spam.

For what it's worth, there are modern LLM detectors with extremely low false-positive rates. The tech has advanced quite a bit since the ZeroGPT days. Personally I've gotten very good results from Pangram Labs. Still can't directly ban people though because false positives are always possible.

CARROT has this and it’s amazing! You can “time travel” back as far as you want. Absurdly far, even. I can tell you that it was 20 degrees in my town on Jan 1st, 1940.

same, i run quite a few forked services on my homelab. it's nice to be able to add weird niche features that only i would want. so far, LLMs have been easily able to manage the merge conflicts and issues that can arise.

Thank you! That's a fascinating paper.

Looking through the poison you linked, how is it generated? It's interesting in that it seems very similar to real data, unlike previous (and very obvious) markov chain garbage text approaches.

I'm genuinely haunted by these TurboTax ads, I see that download app popup at least 3 times a day when I use Apple News. Truly cannot believe someone at Apple though that was an acceptable user experience for ads.

Pangram has extremely high accuracy. While there's no way to prove AI use, it's a very good proxy for that metric. It's obvious to my eyes that the article is written with AI, I supplied Pangram as a citation to convince people such as yourself who didn't notice the AI usage when reading the article.

I actually think AI-human collaboration is quite beneficial. I have a more fundamental issue that it's just bad writing when you use pure LLM generated text. My general feeling is "why should you expect me to spend my time reading something that you didn't care enough to spend your time writing?"

Also, most of the suggestions provided in the AI generated section are just useless. While I think this law is terrible, the suggestions provided completely contradict what the lawmakers are intending. I'll explain what I mean with some of the suggestions provided.

> Narrow the Scope to Intent, Not the Tool

This is essentially a suggestion to throw out the entire law as written. Sure, but this is meaningless advice to lawmakers.

> Drop Mandatory File Scanning

This is the same suggestion as before but rephrased.

> Exempt Open-Source and Offline Toolchains

This is asking them to create a massive loophole in their own law making it useless. Once again, essentially just asking them to throw out the entire law.

> Add safe harbor for sellers and educators who don’t modify equipment or participate in unlawful manufacture.

Two fundamentally different concepts here jammed into one idea. Do you want to add safe harbor for sellers who don't modify equipment or do you want to throw out the entire law and have it not apply to anybody who doesn't participate in unlawful manufacture? These are very different ideas, it makes no sense to treat them as one cohesive concept.

All of these are signals that not much thought went into this. If a human had used AI for ideas and writing assistance, but participated in the writing process as an active contributor, I think they would have caught things like this. I don't think they would have chosen to make multiple bullet points semantically identical. I think they would have chosen to actually cite specific aspects of the law and propose concrete solutions.

Another example, one of their suggestions is to improve the working groups to add specific members. Genuinely a fairly good idea. Having actually read the law, I would have cited the specific passage, which requires that the working group "SHALL INCLUDE EXPERTS IN ADDITIVE MANUFACTURING TECHNOLOGY, ARTIFICIAL INTELLIGENCE AND DIGITAL SECURITY, FIREARMS REGULATION, PUBLIC SAFETY, CONSUMER PRODUCT SAFETY, AND ANY OTHER RELEVANT DISCIPLINES DETERMINED BY THE DIVISION TO BE NECESSARY TO PERFORM THE FUNCTIONS PRESCRIBED HEREIN." I would question, who do they consider to be experts in additive manufacturing? Why does it seem that the working group will be far more heavily weighed towards policy experts as opposed to 3D printing experts? The article suggests that "standards will default to large vendors" yet there is no evidence here that vendors will be included at all.

I can't directly attest to that (never bought an illegal gun) but from my understanding, yes, people have no challenge obtaining illegal guns.

However, you really don't even need to do that. You could just drive across the NY border to a state with looser gun laws, buy one there, shave off the serial number, and bring it back to NY. You could also just steal a gun from one of the many Americans who already own one.

You can also legally buy an unfinished lower receiver in many states (the part of a gun that is typically serialized). Since it's technically unfinished, it doesn't require a serial number. Then you drill a few holes into it and assemble it with off the shelf, also un-serialized gun parts.

Second half of this article has signs of AI slop, as confirmed by Pangram:

https://i.imgur.com/gGIAApA.png

Hard to trust an article like this when the legal analysis and suggestions are being outsourced to an LLM.

> Is that true in New York? Maybe it currently requires permits

What are you referring to as "it" here? When OP mentioned getting a gun from "off the street", that's referring to obtaining one illegally, without a provenance chain or any permitting.

If you want to shoot a CEO, its far easier to buy an untraceable gun on the streets (or obtain a non-serialized 80% lower receiver that you drill yourself) rather than an unreliable fully 3D-printed gun.

It wouldn't protect against this attack though. The Notepad++ update servers were hijacked. Presumably you would allow Notepad++ updates through Little Snitch so you would be equally as vulnerable.

I've seen that as well. I think its still valuable to point out that the text feels like LLM text, so that the person can understand how they are coming across. IMO a better solution is to use a translation tool rather than processing discussions through a general-purpose LLM.

But agreed, to me the primary concern is that there's no disclosure, so it's impossible to know if you're talking to a human using an LLM translator, or just wasting your time talking to an LLM.

Honestly, I can't point out some specific giveaway, but if you've interacted with LLMs enough you can simply tell. It's kinda like recognizing someones voice.

One way of describing it is that I've heard the exact same argument/paragraph structure and sentence structure many times with different words swapped in. When you see this in almost every sentence, it becomes a lot more obvious. Similar to how if you read a huge amount of one author, you will likely be able to pick their work out of a lineup. Having read hundreds of thousands of words of LLM generated text, I have a strong understanding of the ChatGPT style of writing.

What's frustrating is the author's comments here in this thread are clearly LLM text as well. Why even bother to have a conversation if our replies are just being piped into ChatGPT??

Do you really need an LLM to talk on HN? Genuinely, this research seems cool but its hard to trust your findings when there's clearly AI being used heavily in writing the article and in your comments here.

sure, but plenty of software already exists for those devices to block adult content and social media. it works just fine without a header. its actually even better, because that software can even block nefarious websites that would never comply with adding a header

Really big fan of the design of the header on this blog, cool way to represent tags and article information without it being monotonous.

Uh oh, not good that a major Nvidia competitor with genuine alternative technology will no longer be competing... Chances this tech gets killed post-acquisition?

It's wiped from the front page already

It seems to work for me? https://archive.ph/sr0sd

I have a feeling this will get DMCA-ed off of Internet Archive in an attempt to suppress it. Here's the infohash of the archive.org torrent download for future reference, this should allow the file to be retrieved in any torrent client as long as someone in the world is seeding it still.

8105370ed7dba50dc7ec659fd67550569b4dd8a0

I believe they're largely targeting foreign companies who don't care much about US copyright law.

related to your letterboxd suggestion, https://couchmoney.tv is quite good! it uses trakt instead of letterboxd but it's given me quite a few good suggestions. their FAQ describes a similar approach to what you've been talking about, it tries to find movies and tv you like disproportionately like.

Definitely, but Siri is terrible at understanding quiet speech and you have to hold your phone/watch right up to your mouth. A ring is definitely a much nicer form factor for that.