Linux user, dev, bogan The original 69x programmer
A Land Down Under
Joined December 2023
- Tweets10.7K
- Following610
- Followers259
- Likes2.3K
fark, this the second KVM escape in a few MONTHS
Yet another company claiming they can build a foolproof AI detector
Spoiler: it's not possible to make one that's nearly reliable enough to be useful
Researchers have developed a detector that aims to tell the difference between ideas made by humans and those conjured by artificial intelligence
go.nature.com/4ysTHsa
If you think AI is so great
Find ONE online business worth anything at all that lets AI anywhere NEAR their production infrastructure
One of the things not talked about much is how one of the features of art seems to have historically been to convey the work put into the piece, either the piece itself or the practice required to be able to make it
AI is making this more and more difficult, at least online
I think it's perfectly legitimate to hate AI, and AI-generated "art" and such, simply because it is so soulless.
It's perfectly normal to feel insulted when someone throws something at you that they took no care in "creating", and then expects you to spend time ingesting it
The replies to this cover just about everything 😆️
So why are (some) people raving about vibe coding then?
Maybe it's better with guidance
But apparently it needs guidance for everything
Is there a way for a program to ask the OS for a contiguous chunk of space on a disk? Basically reserve the space for a file so it can be accessed efficiently
I know the OS tries to do this anyway but what if you want a guarantee?
(Yes I know about partitions btw)
What do enterprises do these days when they want consistent performance out of their storage?
The performance of SSDs in particular can vary by a factor of 10 or more, particularly on writes
Is there a guaranteed minimum? If so just how low is it?
This is not because Anthropic think they have created something concious
This (yet again) a calculated move
It conveys the idea that their AI IS concious, or at least has some human-like value or worth
Whether they will enforce it at all is another matter
Fun fact: the hallucination rate of LLMs is not evenly distributed
It depends on the subject matter
In particular how commonly known (as expressed in its training data) the fact is
The more obscure, the more likely it is to make shit up
AND the harder it will be to validate