I think my plan of never letting an LLM access the open Internet wins yet again. Maybe GET requests only, no headers or parameters, ask me for every new domain, and only URLs comprising words/numbers no alphanumerics. Maybe. But it seems trivial to allow for a prompt injected attack vector either way
I read it as being more specific than one overarching "position". Long-range strike drones are arguably "offensive action" even in a clearly defensive war, interceptor drones are clearly "defensive action". Midrange strike against targets in occupied Ukrainian territory are a bit of a grey area.
Of course, Starlink/Starshield is already unavailable for long-range strike due to geofencing, and the speed limit was clobbering interceptors right when they needed that speed to cope with the new jet-powered Gerans, so it's still a daft stance overall.
Ukraine's position has been that starlink is used exclusively for drone defense, i.e., interceptors. That is sensible because that's within the contracted terms. However, there is intentional obfuscation of starlink and starshield in all of this. Ukraine also claims that it doesn't use star-anything for offensive drone attacks, which is kind of unbelievable. The contention started when the White House told Ukraine that refineries were off limits, and then Ukraine struck refineries anyway. It seems clear to me that 1) they are using the technology, however it is branded, to reach and possibly cross the border, and 2) this was a signal to abide by the temporary restrictions, as the tri-lateral talks are being negotiated.
The white house agreed to terms with Russia without involving Ukraine. Russia bombed ukraines energy infrastructure literally within hours of Trump saying that the ceasefire was effective and Ukraine waited three days before bombing energy infrastructure in Russia again.
Trump and Musk are being petty because they can't control a foreign nation with a fake agreement that's in the middle of a war. Tried to twist Ukraine's arm. Disgusting shameful behavior. I hope they rot in jail.
I haven't looked into this one but I've seen analysis about past reports that this has more to do with context pollution than any nefarious backdoor. If you pump an LLM prompt with irrelevant information it will just do worse. Some bits of irrelevant information will have a bigger impact than others, but in general it is a good practice to keep prompts detailed but relevant.
Why would you create an AI Trojan that relied on you telling it that you were a Sagittarius? Why are you talking about astrology or nationality when you're programming? I guess
> To surface trojan intentions in open models released with possibly hidden intention in mind.
Phase-change memory is the most advanced high-endurance nvram, with Intel and Micron's XPoint being the most advanced to have hit production. If it had stuck around long enough for the RAMpocalypse, they'd probably sell out their production capacity.
It is the future, and has been the future for almost 40 years already! I bet that's what the hoverboards will use inside their microcontrollers — gotta drive those portable fusion reactors reliably!
From what I've gathered, and I'm still feeling like I'm missing something, most people don't seem to be doing that. They seem to pass the binary in, say to decompile it, and out the other side is something that looks similar but not exact. https://youtu.be/qFEuYUfauNY .
On the recent WAN show LTT said they tried the recent Photoshop clones and they were buggy, feature missing messes. If they were truly reverse engineer decompiles (minus needing a subscription) they would work just as well. But they don't so they're not.
the adobe suite clones are vibe coded clean room implementations, decompiles on the other hand should compile 1:1 to the existing software, and we have ways to verify that's the case. recompiles from my understanding of the term are somewhere inbetween.
I can't speak for others, but I do know that my screen time is filled with this news feed and engineering or other STEM videos. I could never study, it just didn't work for me. But I have a good memory so watching videos tangential to my field helped. I ended up with straight As in my last years of uni, supported by 40+ hours/week of YouTube. Still not sure if that's accounting for 2x speed.
There's a lot of nuance to this. They should consider what we're "staring" at. Is it AI slop content like our parents seem to enjoy? Or more than that? Likewise for social media: maybe consider why in a society where it's prohibitively expensive to do anything and there's a loss of third spaces people are turning to online connections. And beyond that, connections to people from different backgrounds and locations. I can talk to somebody from the other side of the planet online without effort. We can learn of eachothers' hardships and experiences instantly. I think this got shown very well when rednote happened and American + Chinese teens/young adults mixed together en masse for the first time.
I worry that as a smart person you are too optimistic about what your fellow youths are staring at. TikTok and its copycat short vertical video sites are primarily full of content designed to make you dumber, whether it’s prurient content, simplistic and stupid political/cultural takes, or fads. And it of course comes with a joke of a promise that if you waste your own time making the same garbage, then you too can have a career as an “influencer”! I don’t think any of this is giving the benefits your YouTube viewing has.
Other people have mentioned this to me — that Gen Z and younger millennials are likely to fix you in a stare when answering a question.
Bit of a rant incoming:
This is a Gen X teaching failure. As kids we were taught to face front and look at the teacher when they were teaching, and in the era of fears of poorer attention, "attention-focussed" teachers have mistakenly extended this idea that looking the teacher in the eyes when answering is better, when it is worse.
The teacher thinks they have the pupil's attention: they do! But the pupil is supposed to be answering the question: that is where their attention should be.
Kids who do now have trouble directing their attention are not going to do better at answering the question this way, because they are wasting it. As long as the kid is making an attempt to answer the question, teachers should let them stare at the ceiling or a blank wall.
reply