You type in the question or use your voice and it [AI] gives you a detailed answer, like ‘How can I build a bomb?’ and then it tells you how. It is like a human robot! We used it a lot.
I’m pretty skeptical reading this bit. I’ve seen uncensored or jailbroken LLM replies to these kind of questions, they are never actionable, don’t say anything Wikipedia doesn’t, and are hard to provoke if you’re not using an uncensored model.
I have no doubt terrorists are aided by LLMs in a general sense, but am skeptical of any claim that they are providing some material embargoed knowledge that isn’t available elsewhere, in a way that either improves efficiency or effectiveness of their activities, and would want to see real evidence, not an interview snippet.
It's never been particularly difficult to discern how to assemble a bomb, or C4, or napalm, or.... etc. Difficulty in accessibility of violence has never been what protected society. Except, I'm willing to bet, in FBI funding meetings.
Has it become harder to find that information without setting off the fed’s alarm bells?
Was fascinated to learn the PRISM news reduced traffic to privacy-relevant Wikipedia articles, a chilling effect in that case, but indicates tech-savvy folks worry about doing anything on clearnet.
…then again since they’re using CLOUD models, guess my comment doesn’t make much sense…
99.99% of the people consuming that content are law abiding (relatively), "normal" citizens, including probably half of 13 year old boys. I'm sure the feds have access to the "list", but it is virtually useless until after the fact when they can tell the jury "and you see here, they looked it up!"
I read stuff like this and think I must be an idiot because I'm so bad at circumventing the AI safety for fairly benign queries. And here you have folks making bombs...?
This is covered in the full PDF; they have many accounts they spread the queries over and structure them like they're asking for help writing a movie script.
Just for reference, that hasn’t worked for years (the interviews say 2024-25 I think, that kind of attack was patched very early in all the mainstream models) and when it did, you would get bullet point lists GPT 3.5 Turbo style
- first research methods for building effective explosives
- next, assemble the necessary materials to make the bomb
Making explosives is generally fully legal in the US, so IDK if there would even be safeguards for US hosted AI, since there's no real legal issue with doing it. Basically no federal regulations for non-commercial production so long as it isn't stored or moved anywhere, you can literally buy tannerite off the shelf in a sporting good store, "synthesize" it by mixing it and then blow up a huge bomb legally, no license required. YouTube is plastered with people inside the USA making TNT and other materials and then blowing them up.
I can believe some would, but some probably wouldn't care. My local ranch store certainly will happily sell anyone tannerite without even a background check or any sort of scrutiny, you can cash and carry it. Walmart won't but the point being as long as it's legal there will be a "ranch store" that carries it.
>The whole point is not to allow people to make bombs
I mean even YouTube allows bomb making videos and they won't even usually allow videos of people making guns. It's just not very regulated in the US enough to make most companies care. Alphabet Inc. for instance clearly doesn't seem to give a single shit about public access to explosives information, even after the feds subpoenaed Alphabet for Ashley Dugan's Youtube information they still kept his TNT and other explosives synthesis up.
Of course, if you'll allow me to goomba fallacy for a moment, we're supposed to suspend the common HN wisdom here that companies will do anything for a profit / not care unless it costs them something, and also believe that big tech is going to go out of their way to censor the public domain patents they're already hosting on their servers.
AI safety is really brand safety. They don't want to see any more headlines like "You won't believe what OpenAI's chatbot told me!", which was all the rage early on.
> they are never actionable, don’t say anything Wikipedia doesn’t
Researching your way through Wikipedia and the likes certainly counts as "Western education", which as we all know is forbidden by their name. Having an agent read the forbidden stuff for them is just the loophole they needed!
There's lots of knowledge out there about stuff like this. Milennia of humans tinkering with things that go boom. Surfacing it more easily has value (in a manner of speaking; as the @dril tweet goes, "you do not, under any circumstances, 'gotta hand it to them'").
> am skeptical of any claim that they are providing some material embargoed knowledge that isn’t available elsewhere, in a way that either improves efficiency or effectiveness of their activities,
This will sound like a hot take, but consider that terrorists are for the most part, stupid idiots. All the information they need is in books and old patents and what-not, but they absolutely will not have as much success in synthesizing that into effective plans and well-made weapons without having a helpful and patient AI agent to guide them, as they will with that assist.
If the terrorists were very smart, they'd realize that their religion is stupid, that their leaders were mostly corrupt (or themselves stupid), and they'd also probably find something more productive to do with their time.
And obviously yes there are exceptions, since we can all think of infamous terrorist plots which succeeded due to clearly some sophisticated planning and hard work.
> they absolutely will not have as much success in synthesizing that into effective plans and well-made weapons without having a helpful and patient AI agent to guide them, as they will with that assist.
This is demonstrably false.
ISIL’s explosives training material in particular is produced with great professionalism and is of extremely high quality.
Al Qaeda’s was a bit more hit an miss, some of it was mostly ideological propaganda and some of it was outstanding but it was otherwise well structured with clear learning outcomes and assessment activities.
Most of this material was produced well before AI was a thing.
> This will sound like a hot take, but consider that terrorists are for the most part, stupid idiots.
What even is a terrorist?
If your definition of terrorist is "person on the news involved in some FBI entrapment scheme", then yeah they're probably not that bright.
But more generally, terrorists are probably pretty hard to define (one persons terrorist is another persons freedom fighter, etc), and I would imagine include a whole range of intellectual capacities.
> "one persons terrorist is another persons freedom fighter"
yes, but only to weak-minded cultural relativists.
I don't mind stating a definition explicitly. If you would consider killing nonviolent civilians, merely to send a message to others or to prove yourself to your "god," you're a terrorist.
Note: The only real complexity in the definition is that during war, embedding military equipment among civilians happens sometimes, and when it does, it's the ones choosing to do that, war criminals by definition, who bear the responsibility for those innocent deaths when those locations then become legitimate military targets.
I'll omit giving examples to avoid triggering those who like to "bothsides" some of the more well-known terrorist groups lately.
> If you would consider killing nonviolent civilians, merely to send a message to others or to prove yourself to your "god," you're a terrorist.
The US is a "christian" nation which has been doing this for decades, even centuries at this point. As recently as a few weeks ago and the official government seems to be pretty joyous about it. I'm pretty sure it's the same "god". Not sure who this god guy is but he seems like an asshole who can justify anything.
> embedding military equipment among civilians happens sometimes
"happens sometimes" is in the passive voice. What if you live in a war zone and the war isn't far away? Are the people stockpiling AR-15's in the US legitimate military targets? There's a fun 80's movie about Cuba invading the US that you should watch some time. Is that "pro terrorist propaganda"?
Boko Haram is not a mentally disabled person tricked by the FBI they are real terrorists who attack schools to stop children from learning. They're the epitome of stupid and backwards.
> If your definition of terrorist is "person on the news involved in some FBI entrapment scheme", then yeah they're probably not that bright.
That's a great example of selection bias: if they were intelligent enough to be capable of causing harm at that scale, they would have immediately realized the new guy suggesting targets and offering to buy them explosives was an FBI agent.
I'm not admiring or excusing anyone. I'm saying you probably find a whole range of IQs among them. Because why wouldn't you? Whose to say if the userbase of this site grew up in the wrong conditions, a portion wouldn't be terrorists too?
>If the terrorists were very smart, they'd realize that their religion is stupid, that their leaders were mostly corrupt (or themselves stupid), and they'd also probably find something more productive to do with their time.
In my head I was running through a whole lot groups to replace "terrorists" with in that sentence...
> If the terrorists were very smart, they'd realize that...
There are a lot of very smart people joining cults, including people who just wanted to study them and were fully aware that what they believed in was nonsense. Terrorist groups are cults.
Just because you consider yourself smart because you have a PhD, high IQ, built a successful company or practice some intellectual activity and are good at it doesn't mean you are immune to the manipulation tactics used by cults. Once sufficiently brainwashed, your judgment will get clouded, but it doesn't mean you will lose your technical skills, which you will put to the service of the cult.
Think about Nazi Germany. Germans weren't idiots, the simple fact that most of the world had to turn against them to defeat them is proof enough. But they still managed to get manipulated into committing the worst crimes.
The reason it worked is not stupidity, it is misery. They were at their lowest following WW1, and Hitler promised them better. Same idea for terrorist groups and other cults, people who are experiencing hardships are the easiest to indoctrinate, and there is no shortage of such people in Nigeria.
Well said. As mentioned in another post, it's not about intelligence, it's about vulnerability. One particular type of misery is mourning, and the western world does a pretty good job of creating mourners in a number of places where terrorist groups flourish.
His videos of making them were on YouTube for years, publicly. It's legal to synthesize the explosives he made. What they did was charge him for the first amendment protected activity that a terrorist then found, and then they claimed that because he made some money because a few people donated a small amount to him for making the videos, and thus he needed a commercial license for being in the business of making explosives.
By the way this is the same thing they tried to charge FPSRussia (the first time, before they convicted him for weed) for and failed.
You didn't read what I said. Try reading the law instead, or even the charge. It's for being in the business of manufacturing explosives without a license. A license isn't needed to manufacture explosives. One is needed to manufacture them as a business venture. They are claiming since he got a little money from Youtube or viewers he was in the business and that was illegal.
This failed when they tried it with FPSRussia.
====== re: below due to throttling ======
>I think "I make explosives for YouTube revenue" falls squarely within the business territory.
Different than what you said initially which was merely making them, which is why I clarified.
>> Licensed manufacturer. A manufacturer licensed under this part to engage in the business of manufacturing explosive materials for purposes of sale or distribution or for his own use.
engaging in business of your own use is not same as non-business of your own use. It's not uncommon for a business to use explosives for their own use as course of operations. It is also not uncommon for people to synthesize and use tannerite recreationally without a license, legally, for non-business use.
>I did read what you said; that's why I quoted part of it.
If you read it then you know you maliciously selectively quoted it then. If that were the end of it and that made it legal, I would have stopped there, but you cut it off there because it was more convenient to your rebuttal to ignore the rest. I only thought you had not read it, because I was being charitable to try and assume good faith.
> Licensed manufacturer. A manufacturer licensed under this part to engage in the business of manufacturing explosive materials for purposes of sale or distribution or for his own use.
> The presence of a video on YouTube is, quite simply, not evidence of the behavior in it being legal.
It is evidence of looser temporal relationship to the timing of the NOLA bombing, leaning more towards it released as general education rather than prepared for the NOLA bomber. Hence, supporting but not direct evidence of the behavior not violating the law regarding intentional instruction of terrorists. Therefore I assert it is relevant to the legality.
It's only when you maliciously quote it out of context that you can misrepresent my argument to be the sole fact that it was on youtube for years means it is legal. The fact it's not sole direct evidence of being legal doesn't mean it's not useful accompanying information.
>I quoted one part that was just deeply goofy logic.
It was only deeply goofy logic when you isolated and then straw manned what I'd claimed.
If you feel the need to save face, then sure. Simply saying "it was on youtube for years" does not mean something is legal. Though I don't think that was ever in contention.
I have no doubt terrorists are aided by LLMs in a general sense, but am skeptical of any claim that they are providing some material embargoed knowledge that isn’t available elsewhere, in a way that either improves efficiency or effectiveness of their activities, and would want to see real evidence, not an interview snippet.