cvhc

8 ms·

61.
by cvhc 1y ago
(1) One would prefer precision over recall for spam detection, so bear some false negatives; (2) these classifiers are subject to adversarial attacks -- you may see spammers repeating the same tricks to circumvent the detector for a while a
62.
by cvhc 1y ago
Wow. I love this. As a not very social language learner, I've used LLM to generate everyday conversation for me to practice. This tool streamlines this experience. I don't think Google will continue the development. It'd be g
63.
by cvhc 2y ago
I still use Google a lot. I don't bother chatting when a few keywords will bring me to the resource I want (90%+ of my searches). Besides, Google has some convenient features that I frequently use, e.g., currency/unit/timezon
64.
by cvhc 2y ago
The format is clearer than podman generate systemd or kubernetes YAML. And the integration with systemd is great. What annoys me is Podman upstream doesn't offer a repo for Debian/Ubuntu. I was stuck at version 4.3.1 on Debian sta
65.
by cvhc 2y ago
My friend why so irritated... I was just explaining why you two got different numbers. And my numbers of other companies are also calculated from the start of 2022. I guess my broken English doesn't match today's bot quality :)
66.
by cvhc 2y ago
Agree. A majority of people on HN are in a startup mood so they feel a company should market aggressively to attract investments and expand. But I don't think Google would achieve more than marginal gain were they to aggressively make
67.
by cvhc 2y ago
This really depends on which day of 2022 you start the calculation. But to be fair, you can claim the same for AAPL (+25%)/MSFT (+22%)/AMZN (+22%)/... It's just the up and down of the entire market (and these big techs d
68.
by cvhc 2y ago
> Possibly by trying to find a business model that can support Chrome development just like all other Chromium (and non-Chromium) based browsers? What would this business model be like, if, say, Google Chrome is eliminated? As a referenc
69.
by cvhc 2y ago
I've been using maddy. It seems this has more features (like web interface, reputation tracking etc).
70.
by cvhc 2y ago
I was going to say the same. I wonder what exactly o3 costs. Does it still spend a terrible amount of time thinking, despite being finetuned to the dataset?
71.
by cvhc 2y ago
And a stone tablet, not a tablet computer. As an ESL speaker, my first impression was "a tablet computer displaying unknown language found in Georgia, US". Feels like some extraterrestrial technology.
72.
by cvhc 2y ago
Many Americans don't realize how easy it is to fall into some sort of "illegal" status in the US. I have been stranding in the US with expired status at the beginning of the pandemic. I filed change of status but USCIS never
73.
by cvhc 2y ago
While in this case distros include cups-browsed maybe as a feature, I always feel it's a bad thing Ubuntu/Debian (and maybe all deb-based distros?) automatically bring up almost all services upon installation. This means you can i
74.
by cvhc 2y ago
I just feel SELinux would add too much burden to sysadmins. I use CentOS + SELinux in one of my VPS and it's already painful. I've been sysadmin in university labs for some years. I did what I think is reasonable, setup a firewall
75.
by cvhc 2y ago
For 1, tbh I don't know the rules. technically it was a regular model and can store some data. Or maybe they simply suspected it's a usb disk. And 2 was true. But it was after weeks, and of course I didn't wait until then to
76.
by cvhc 2y ago
Another side. I always worry more about being locked out by 2FA (a main use case of non-discoverable FIDO keys) as a consequence of lost/retired-then-forgotten keys. This happened once when the US custom detained my electronics, includ
77.
by cvhc 2y ago
In the FIDO2 case, only the derived keys are extracted. The master key that derives non-resident keys isn't extracted. So I think it's not possible to really copy the key. In the cases of FIDO2 resident keys (passkey) / PIV &
78.
by cvhc 2y ago
There is one difference between gibberish Chinese and Latin character sequences. In Chinese, each character indeed carry some meanings (like a word). So I guess the model may hallucinate some output inspired by these meanings. In the case &
79.
by cvhc 2y ago
Right, completely gibberish. as a native speaker, I can recognize at most 4 characters, and not even one subsequence makes any sense. Actually just by shuffling these characters you have a good chance to get some specious translations (addi
80.
by cvhc 2y ago
An article from The Register, which also questioned the claim: https://www.theregister.com/2024/08/13/trump_musk_livestream...
81.
by cvhc 2y ago
I have been trading in my phone every year to get the new Samsung S2x (now S24) model since S21. There were always new model deals plus some regular cashback. I ended up paying <200 USD every year. It's like a manual phone upgrade p
82.
by cvhc 2y ago
I wonder why the top level has to be an object instead of an array... I have some pretty normal use cases where I expect the model to extract a list of objects from the text. ``` openai.BadRequestError: Error code: 400 - {'error':
83.
by cvhc 2y ago
FOF ¶ 334. Google’s default placements on Firefox generate 80% of Mozilla’s overall operating revenue, demonstrating that the vast majority of query volume on Firefox goes through defaults. Ref: https://s3.documentcloud.org/
84.
by cvhc 2y ago
"President Trump", "Donald Trump" complete normally, and suggest "45th US President". I think it had something to do with the recent attempted assassination. I tried a few related phrases and it won't comp
85.
Show HN: GPG-fingerprint-filter-GPU – Collide GPG key fingerprints, CUDA powered(github.com)
6 points by cvhc 4y ago|0 comments