usef-

7 ms·

91.
by usef- 2mo ago
Plenty of providers on that list aren't fly-by-nights and have easily proven themselves in past models. Jeremy Howard does demos on Fireworks. A nice thing about open router is you can specify filters, like "US hosting without dat
92.
by usef- 2mo ago
Who's suggesting that? They're sending reports to projects to fix. It's up to the projects on how they fix them. No, I don't think all bugs are fixed. The point of the project (glasswing etc) was to fix as many as possib
93.
by usef- 2mo ago
I do think security issues in core building blocks like curl and the linux kernel (and almost every significant project) are still a concern even if developers are being sloppy on newly-built apps. This isn't an "are LLMs net good
94.
by usef- 2mo ago
This seems a highly controversial post for some reason, judging by the repeated downvotes/upvotes. I'd be curious what I got wrong.
95.
by usef- 2mo ago
There are many possible international producers, it doesn't need to be physically transported, every household could near-instantly have a copy, it puts local business at a disadvantage if they can't use the same tools as internat
96.
by usef- 2mo ago
It does seem inconsistent that we currently ban closed models that fail the safety tests but not the open. I feel like the only consistent position is to either care about the safety issues (like people producing biological weapons) for all
97.
by usef- 2mo ago
Banning software/data from being put on the internet vs banning GPU hardware is fundamentally different though, isn't it?
98.
by usef- 2mo ago
Yes, to parse logs afterwards and understand, it wasn't active defence from what I've heard? Definitely embarrassing for the closed vendors though (they've since added hugging face as a trusted vendor)
99.
by usef- 2mo ago
Distillation of this kind isn't the only data that is fed into models, it's still a small minority. It might be weighed higher in training to learn the thinking patterns/etc but you need a lot more tokens.
100.
by usef- 2mo ago
They're not slop, if you're talking about recent ones. You might still be operating on information for a year or two ago. Here's the curl project talking about the strain they're under from real reports (despite being a
101.
by usef- 2mo ago
They are doing that (see their project glasswing over the past few months), but there's a lot more code in the world than you realise. The problem with rolling it out is that bad and good actors can both use it at the same time, and
102.
by usef- 2mo ago
Are you claiming that LLMs aren't finding new issues compared to previous methods? There's a huge number of security issues coming out in recent months, especially via Anthropic (glasswing etc). We don't have to take their wo
103.
by usef- 2mo ago
Has it impacted your use case negatively?
104.
by usef- 2mo ago
I very much doubt they would want to abandon the wider community. They do benefit from others using it. To the contrary, I think any problem with this release would be jumped on harshly so they're being more careful than usual. There&
105.
by usef- 2mo ago
They also have made a release, in a sense, since they moved all Claude code users to it a month ago. (And apparently no one noticed). I think they're taking things "gradually" as they are under a lot of scrutiny and no rush f
106.
by usef- 2mo ago
The typical ram has surprisingly not increased very much in 10 years. > April 2016, 8 GB was standard across the 13-inch MacBook Air range ... Now it's 16. Rich nerds will have quite a bit more. But I suspect the standard of model
107.
by usef- 2mo ago
If your workload fits long batches throughout the night you could schedule them better, yeah. But I think very few have a usage pattern like that? Given the hardware shortage in the world, I suspect renting ("sharing") via APIs wi
108.
by usef- 2mo ago
In previous years Sam's huge hardware deals were seen as a crazy example of the bubble but currently are looking rather fortunate. Though Dario's reasoning still applies this year, as the scale keeps increasing. Very interesting t
109.
by usef- 2mo ago
On the plus side, the competition seems a bit lousy lately
110.
by usef- 2mo ago
I'm curious if this is just to pressure Google into separating their crawlers
111.
by usef- 2mo ago
How are they playing both sides? I thought their scraping products were also about having it behave and not take down systems
112.
by usef- 2mo ago
I did, hah, edited. I did not know that word.
113.
by usef- 2mo ago
The article is basically saying the opposite: models don't need you to specify as much now. None of it is talking about better or precise language, it's about what you should say to it. (I love how often the highest-voted commen
114.
by usef- 2mo ago
I don't think most of the layoffs that have happened so far are about present-day AI. It's often an easy excuse though.
115.
by usef- 2mo ago
Or the amount of software increases tremendously
116.
by usef- 2mo ago
fwiw, I don't think it was deliberately provocative for the goal of "attracting attention", it seemed to be his genuine belief. I remember feeling similarly after seeing so many people denying what had already changed.
117.
by usef- 3mo ago
They don't have a track record of benchmaxxing. The simplest explanation is that the blog post lists the things this model is better than Fable at, but not the things it isn't.
118.
by usef- 3mo ago
Yes, all chat interfaces store the history as it's part of the UI promise (unless you open an incognito chat) and is fully server side. When people talk about retention they mean API usage and terminal agents, which run on your device.
119.
by usef- 3mo ago
People _hate_ when they talk about safety. It doesn't seem like it's helping them.
120.
by usef- 3mo ago
The same is true in every realm: taste matters. But that's not a reason to blanket-discourage expression. (I do wonder if the people here who hate it are picturing students who blindly highlight pages, and the people who like it are re

More ›