Grimblewald

3 ms·

31.
by Grimblewald 2mo ago
the back button hijack on mobile is a nice touch.
32.
by Grimblewald 2mo ago
This is a terrible idea. It sounds like the way they're going to do this is by looking at specific word choices, which is the standard say AI "detection" currently works. The problem is that if it is "undetectable"
33.
by Grimblewald 2mo ago
i tend to give ai a "voice" file, which it must use to write. I hope this doesnt spread to providers i do use, since if it does i mugbt just forgoe ai, i cannot stand ai prose.
34.
by Grimblewald 2mo ago
well, i's love to see what lead to mythos trying this. if instructed to do it, which it likely was, it doesnt prove much of anything.
35.
by Grimblewald 2mo ago
or maybe SO is back to its community sans clout chasers and tourists? If a vaninishing minority drive a given community in content generation and discourse, then lurkers etc leaving isnt as meaningful. you could lose 99% of users on many pl
36.
by Grimblewald 2mo ago
automatic coding is a flow thing with 3.6, but I agree it isnt fully there, it does need specific per function instructions to maintain coherence, with a few exceptions Im unsure hold in general which I got via an experimental harness.
37.
by Grimblewald 2mo ago
While I agree the claims may be strong, they're not inconceivable. I still think current AI architectures are fundamentally incapable of replacing us in any of the ways discussed, however it is now something I consider plausible as see
38.
by Grimblewald 3mo ago
dunno what your interface looks like, but no. not avaliable to me save via openrouter or similar, but the fool i am i purchased a year of claude sub on the premise i'd get 4.5 or better.
39.
by Grimblewald 3mo ago
mirrors my own findings as well. It royally bungles projects in a way that make git a bigger godsend than it should be (evem though it is). I really miss the 4.5 era, that was a magic time.
40.
by Grimblewald 3mo ago
Please correct me if I'm incorrect, but it seems to me those numbers are describing a fairly different situation to this one. Anthropic serving their own model to their own users at that scale gets cache hit rates and machine utilisati
41.
by Grimblewald 3mo ago
if this qualifies bootleg, point me at non-bootleg frontier ai.
42.
by Grimblewald 3mo ago
as an educator, does match my experience though. I do something similar, tho a little more sophisticated than white text for some assignments so that students not cheating have a way to standout from the pack. in my most recent class of 48
43.
by Grimblewald 3mo ago
"needs safety eval" also anthropic "we're upset were not being considered for military contracts" come on, which is it? Is it all about saftey or is it that only US/Israeli ai is allowed to kill? Seems to me th
44.
by Grimblewald 3mo ago
I largly agree, hence why the prudent move is not visiting shitholes like the (current) USA with anything important on your person. However, if you must do it, there are better options than duress pins that wipe a device.
45.
by Grimblewald 3mo ago
unfortunately one of the main applications of current wave "AI", in the long term, is going to be mass surveillance. Meta is a big player on that front.
46.
by Grimblewald 3mo ago
I could be wrong, but my understanding is that the dummy OS views the remaining space as legitimate and accessible free space. Using dummy directly is of course dangerous, as you might overwrite sectors with legitimate data, but also, you c
47.
by Grimblewald 3mo ago
true, but in that scenario you're going to prison either way. If you legitimately use the dummy for daily driving and hidden for sensitive work, then it's better than nothing. Obviously a good alternative is a dummy device but it
48.
by Grimblewald 3mo ago
It's been massive for me as well. I would say my web developmemt is best described as dabbling in web development, and I mostly do so for websites that cater to academic interests and conference / society managment and organising.
49.
by Grimblewald 3mo ago
given dark patterns employed to date, what makes you think this would generally work? If i wanted to capture all consent-o-matic users, I'd simply use text of buttons that don't line up with what the buttons descriptors are. human
50.
by Grimblewald 3mo ago
VeraCrypt has a cool function which is a reserved space for a decoy OS.[1] Everything else registers as free space while decrypting to dummy volume. You make the dummy volume look lived in, and forget. provide dummy password, volume decrypt
51.
by Grimblewald 3mo ago
I think we largly agree. The point I'm making is that we use technology to solve discomforts, and discomfort tends to be something which would otherwise manifest as evolutionary pressure. This doesn't remove the fact we will chang
52.
by Grimblewald 3mo ago
sure, it can be. However, I'd argue in a large amount of cases it fits. Agriculture, medicine, transport combine massivly lower the bar for base survival let alone reproduction on all skill checks otgerwise place don metabolism, physic
53.
by Grimblewald 3mo ago
also, technology largly exists to remove the discomfort of evolutionary pressure. This doesnt mean the mutations that give rise to evolutionary changes in the population, but it does mean we've removed or mititaged many of the mechanis
54.
by Grimblewald 3mo ago
why not cut to the chase and simply remove user access to devices? preventing non playstore installs and restricting adb are addressing symptoms, user behaviour, not actual security issues. RIP android.
55.
by Grimblewald 3mo ago
alternativly the HF is a gpt2/strawberry/mythos style marketing stunt. Does no one remember the extreme fearmongering around gpt2 which barely produced coherent text?
56.
by Grimblewald 3mo ago
And I have evidence that anthropic distills from openai, moonshot, alibaba etc. So unless anthropic holds itself to the standard they're implying should be followed, why should I care others did to them as they do unto others? Seems li
57.
by Grimblewald 3mo ago
I think google hate is summarised nicely by their removal of the "do no evil" slogan.
58.
by Grimblewald 3mo ago
I get resistsnce. GDPR is good in theory but implementation has been shocking. Dark patterns in banners, banners when not required (ie no relevant cookies) etc. the gdpr banners as a whole make the internet a shitter experience. I categoric
59.
by Grimblewald 3mo ago
for usable local layperson applications, I think qwen models are king. However, at hosted/frontier scales I'd agree for code. However for visual comprehension etc. for me qwen is the top of the line, and it isnt even close. Qwen V
60.
by Grimblewald 3mo ago
Didnt the piss tint come after their big studio ghibli heist?

More ›