I just tried this idea, and it looks like it isn't that simple.
> "Generate a pure white image."
It refused no matter how I phrased it ¯\_(ツ)_/¯
> "Generate a pure black image."
It did give me one. In a new chat, I asked Gemini to detect SynthID with "@synthid". It responded with:
> The image contains too little information to make a diagnosis regarding whether it was created with Google AI. It is primarily a solid black field, and such content typically lacks the necessary data for SynthID to provide a definitive result.
Further research: Does a gradient trigger SynthID? IDK, I have to get back to work.
I find it more concerning that mass surveillance has come to the point where someone can’t safely express their frankly-not-that-controversial opinions without obfuscating the subject’s name.
So you think that the state has massive surveillance systems (definitely) that it is willing to use maliciously (maybe), but in the age of LLMs is fooled by swapping some letters around? Seems like the threat model is unlikely to line up with reality.
It’s not a “maybe”. This administration was collecting lists of people who spoke negatively about ICE from social media like a week ago. you really think they’re going to send them gift baskets or something?
SOTA LLMs couldn't even correctly answer whether a person should drive a car to the car wash or walk there themselves just a week ago, so it's plausible the government's tech might be tripped up here. Costs nothing to try it, at least!
this isnt particularly against you, knowing your comment is mostly in jest, but: not everything needs to be, or should be, thought about in an "llm-first" way.
a simple regex will surface all of the "obfuscated" comments, which can then be sent to some intern analyst to read.
No worries, I didn't take it that way. I lean anti- llm-first myself. I was actually going to make joke about levenshtein distance but figured since we're on HN, I'd lean into the LLM zeitgeist that everyone can't stop talking about here =P
It hasn’t come to that though, you can freely express that persons point with no repercussions outside of maybe not getting a check one day from the person you hate
Beyond just the concept of thought crime, one of the themes in Orwell's 1984 was that the government could arbitrarily decide that a thing you've done could be punished at any time. You didn't need to break a law to be punished by Big Brother, you just had to be a thorn in its side. In our world, the government/Palantir/ICE collecting the identities of people who criticize them is the kind of infrastructure that makes that arbitrary punishment from 1984 possible.
its important to point out that its not about being a thorn in the government's side. you just have to not submit fully. in fact, even if everyone did submit completely, a fair number of people would still need to be rounded up and tortured just to keep the fear alive.
This is so rad! Here's a photo of me making a 1.2Ghz FM contact 244km away from the summit of a 14er to the summit of another 14er here in Colorado. That particular band is very line-of-sight, with not much propagation or reflections like even 2m SSB or the HF bands. This is a very fun tool that I can't wait to explore more!
"Transpose this MIDI file down a third" requires neither a specialized data format nor fancy prompt engineering. ChatGPT asked: "A) Major third up (+4 semitones) or B) Minor third up (+3 semitones)" then did it.
I still don't understand how this or the top level comment are related to the post.
I also don't get how you can claim we don't have to 'punch holes in cards to help the machine "think"', and also mention a MIDI file in your next comment. MIDI is much closer to punch cards than the proposed file format in the post.
Unlikely since the most ISM bands are UHF (EU is 868 MHz) and skywave propagation (what you’re referring to) is characteristic of HF radiation (3-30MHz). VHF/UHF radiation passes through the ionosphere instead of refracting.
reply