nedruod

4 ms·

1.
by nedruod 13d ago
Doesn't look like pi.dev has anything like auto-mode. They tout running in a container, which you should do regardless, but the scope of soundness that has as a security plan is limited. If some information is in the container, and the
2.
by nedruod 14d ago
You assume alignment and marketable are the same. That's not true. You would willingly work with an unaligned model. At best, you might say you wouldn't if you knew, but (a) you might not know, (b) you wouldn't be representat
3.
Every Reward Bends: Security Incentives and the OpenAI Hugging Face Incident(substack.norabble.com)
1 points by nedruod 25d ago|0 comments
4.
The Defender's Window(openai.com)
3 points by nedruod 1mo ago|0 comments
5.
Why It Hasn't Happened Yet: Capable AI and Malicious Intent(substack.norabble.com)
2 points by nedruod 1mo ago|0 comments
6.
Mark Zuckerberg: The Future Is for Everyone(about.fb.com)
4 points by nedruod 2mo ago|2 comments
7.
by nedruod 2mo ago
It's not so absolute. In the first order, yes, if you don't grow generation or transmission, it has to increase prices. But in the second order, those things happen, and the end result can be lower prices on scale. That leaves it
8.
Banning data centers would blow up the U.S. economy(noahpinion.blog)
11 points by nedruod 2mo ago|5 comments
9.
by nedruod 2mo ago
I use them, and the reason is simple - they attract users. All these directions to not use AI here, not use it there, not use it anywhere, they're all basically missing the point, which is what someone was trying to do in the first pla
10.
Nobody Was Watching: Anthropic, OpenAI, and Open Models(substack.norabble.com)
10 points by nedruod 2mo ago|0 comments
11.
by nedruod 2mo ago
The China point cuts both ways though. The main point from that would be that everyone's opinions are irrelevant. No point in supporting open-weight, because China is going to do what China is going to do. No point in supporting less o
12.
by nedruod 2mo ago
If you think I've insinuated they don't have a motivation, you've misread. It's undeniable they have motivations (though I pointed out, the signatories and non-signatories all have the same motivations). From a PR perspe
13.
by nedruod 2mo ago
Ah, the position they don't state. So why respond to this press-release, which doesn't state that, since obviously it adds zero value? And why aren't the companies that signed the letter supporting open-weights, who have the
14.
by nedruod 2mo ago
This take is poorly informed and poorly thought out. It's just a cynical vibe, not a cogent argument. If they wanted to squash the competitors, they'd advocate for a ban on Open-Weight models. How else are they going to prevent th
15.
An OpenAI Model Escaped Its Sandbox. Where Was the Observer?(substack.norabble.com)
2 points by nedruod 2mo ago|0 comments
16.
OpenAI Model Hacks into HuggingFace During Cybersecurity Evaluation(thezvi.substack.com)
3 points by nedruod 2mo ago|1 comments
17.
Tech jobs market in 2026, part 3: hiring managers and job seekers(newsletter.pragmaticengineer.com)
3 points by nedruod 3mo ago|0 comments
18.
Why American data centers can't plug in(worksinprogress.co)
2 points by nedruod 3mo ago|0 comments
19.
by nedruod 3mo ago
In terms of slippery slopes, this argument is climbing atop one. Where does that end? Should we ban doctors from performing surgeries that would save someone's hearing? Should we ban protective gear that might diminish the size of the
20.
by nedruod 3mo ago
Can I mention the irony that this seems a "bit" paranoid? "bit" because, yes, you do have some good examples of failures to call to, but still, consider how much good would have to be chucked to have avoided those by a g
21.
by nedruod 3mo ago
Sure, you can imagine it, but you should be able to imagine better solutions too. Far easier to ask for a government ID, or better yet, have a third party, which is trusted, use a cross-check, so that the raw data is never shared, just a bi
22.
by nedruod 3mo ago
You cannot undo anything that's history, including not testing. If you missed a chance to get other tests or treatment early in life, you don't get the chance to fix that later either. It would be easier to be cautious on penetra
23.
Should every baby's DNA be sequenced?(economist.com)
23 points by nedruod 3mo ago|88 comments
24.
Or Equivalent Experience: Lazy Mistakes in Hiring and the Truth Behind Jobs Data(substack.norabble.com)
2 points by nedruod 4mo ago|0 comments
25.
Challenges for AI Misuse Prevention: Jurisdictions, Open Models, and Privacy(substack.norabble.com)
2 points by nedruod 4mo ago|0 comments
26.
AI Safety Is Underfunded by Design: Model for Incentive-Aligned AI Safety Policy(substack.norabble.com)
1 points by nedruod 4mo ago|0 comments
27.
Adversaries Leverage AI for Vulnerability Exploitation(cloud.google.com)
1 points by nedruod 5mo ago|0 comments
28.
Focus Areas for the Anthropic Institute(anthropic.com)
2 points by nedruod 5mo ago|0 comments
29.
AI Companies Can't Regulate Themselves. They Should Regulate Each Other(lawfaremedia.org)
1 points by nedruod 5mo ago|0 comments
30.
Academics Need to Wake Up on AI, Part III(popularbydesign.org)
3 points by nedruod 5mo ago|0 comments

More ›