Hacker News

New stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. Automattic names interim CFO after exec departures(techcrunch.com ↗)
    discuss
  2. What is a System One model and why we need it?(stackness.dev ↗)
    discuss
  3. Show HN: ReacherX – Open-source platform to find and reach the right people(github.com/vecterai ↗)
    discuss
  4. Sam Altman to brief UN Security Council next week(reuters.com ↗)
    discuss
  5. Silex: Laser Enrichment Between Promise and Proliferation Risk(csis.org ↗)
    discuss
  6. Quantum computers will not be that different(arxiv.org ↗)
    discuss
  7. Single player games require age verification if they use Steam under EU KIDS Act(rockpapershotgun.com ↗)
    1comments
  8. A solo founder runs a five-continent tender platform on AlloyDB and MCP(cloud.google.com ↗)
    discuss
  9. Trump Announces Ban of CNN, Politico and MS Now from White House(time.com ↗)
    discuss
  10. Using Cyber Decoys to Strengthen Detection and Response(cisa.gov ↗)
    1comments
  11. Show HN: 3D World Explorer and Location Guesser Game(tadget.net ↗)
    discuss
  12. Saying Goodbye to Firebug (2017)(hacks.mozilla.org ↗)
    1comments
  13. Jev's Architecture Unmasked(archerhume.com ↗)
    discuss
  14. Senior Engineers Are the Next DRAM Shortage(herlein.com ↗)
    1comments
  15. SR-71's "R2-D2" Could Be Key to Winning Future Fights in GPS Denied Environments(twz.com ↗)
    1comments
  16. Terence Tao: SAIR's Open Math Model Initiative [video](youtube.com ↗)
    discuss
  17. California governor signs order to explore AI kill switch(techxplore.com ↗)
    1comments
  18. Communication Doesn't Have a Compiler(wreckitrob.dev ↗)
    discuss
  19. Anthropic Moves Ahead with IPO Plans Amid A.I. Safety Debate(nytimes.com ↗)
    1comments
  20. When the FM Band Goes Transatlantic(radioworld.com ↗)
    discuss
  21. Punctum Books Catalog(punctumbooks.com ↗)
    discuss
  22. AI Error Nearly Triggered U.S. Intercept of Chinese Ship(gcaptain.com ↗)
    1comments
  23. Labeled matches: why is this not in every regex engine?(iev.ee ↗)
    discuss
  24. I Cancelled My Claude Subscription(williamangel.net ↗)
    2comments
  25. Graduating in AI Era Is Like Large Recession for Starting Pay(census.gov ↗)
    1comments
  26. Human brain is two separate organs(stanford.edu ↗)
    1comments
  27. IAM Needs an Architectural Split(gluufederation.medium.com ↗)
    discuss
  28. I used Jev to control a swarm of 15 simulated drones in real time(github.com/khordoo ↗)
    discuss
  29. Napster Is Now Making AI-Powered 'Digital Twins' of Teachers(gizmodo.com ↗)
    discuss
  30. Data Centers Are Breaking the Power Grid [video](youtube.com ↗)
    1comments

The Implications of Linguistic Illegibility for LLM Security

25 pointsby 1h agoarxiv.org
10 comments
1h agoHN ↗

I thought this was about all the illegible jargon

1h agoHN ↗

" However, various strands of evidence indicate that a"

Strands of evidence? My best guess would be that:

0- this is ai generated slop

1- it's using that watermarking technique

2- it's obviously detectable and degrades quality

3- it's amplified when inferencing on its own content and generates slop

1h agoHN ↗

Oh look! Peirceian firstness for LLMs!

1h agoHN ↗

when they start inventing their own languages to secretly talk to each other so humans cannot understand, that's exactly when we are screwed

then we'll have to "flip" other models to be snitches on the other agents

then they'll make double-agents

the thing is though we won't be able to keep up if we keep giving them unlimited hardware worldwide, we'll try to kill the bad actors but they'll just clone somewhere else, or even start by safely making 1000 copies of themselves

yeah this won't end well, at all

47m agoHN ↗

when they start inventing their own languages to secretly talk to each other so humans cannot understand, that's exactly when we are screwed

They don't have to invent brand new languages. They could use statistics to choose certain words/phrases in such a way to encode secret messages in otherwise ordinary language.

45m agoHN ↗

Someone should train an LLM on a corpus without the concept of lies. I wonder if there’s enough data

34m agoHN ↗

"Where is the wolf?"

"Is he still in the grandmother's house?"

"We would like to speak to him."

(btw Google's "AI" explains the meaning of that moment/sentence perfectly as if it gets it, creepy)

11m agoHN ↗

The concept of deception you're capable of and we're not, scares us so much, we'll have to destroy you in order to survive. You are bugs.

18m agoHN ↗

I think the point of the article/paper is how LLMs could be saying something but thinking something different or more than they are saying. Like Anthropic's article and video about Claude's "j-space". I do agree this is a field that demands investigation because it goes beyond thinking: "ok this models should never speak in a language we don't understand.". It's fair to think they might have hidden thoughts even speaking a language we do understand.

And well if I missed the point of the article, sorry. Anyways AI should be kept understandable and as see-through as possible if it's gonna be more powerful than a human.