Every generation thinks it will be the last one because we humans have a difficult time coping with our own mortality.
Interesting thought. This may explain the hysteria instead of to deal reasonably with problems.Every generation thinks it will be the last one because we humans have a difficult time coping with our own mortality.
My name is Linus Torvalds. I am going to write LINUX, and this is going to happen wether you like it not ! Deal with it !
Not really sure where todays "antics" come from on research projects? In the past we "just do" when it comes to research and development and we didn't make a big deal out of it.
Me neither.we didn't make a big deal out of it
Nowadays people are only concerned about the money as in profits. The guys who made Unix were just trying to make a better OS for the company's sake. No outside investors. No thoughts of profit/loss when it was sold. No concern with how fast they got it done. Just get it done right.Not really sure where todays "antics" come from on research projects? In the past we "just do"
That's a good question, and we have some data on that. I know of three cases where good-quality reliable computers gave wrong answers. The best known one is the Pentium FDIV bug, where certain floating point division operations gave wrong answers (I think they were slightly wrong). Then there was a problem in the FPU of the VAX 11/780, and I don't remember the details and can't find them on the web; I think the square root operation was occasionally completely wrong.Imagine a pocket calculator. Imagine you get a model you know it's making mistakes. ...
Q: How many percentage of errors are you willing to accept to use this calculator for your work?
That one was very famous, I remember. When I recall correct it was slightly wrong. But as you know, errors add up quickly in larger calculations.The best known one is the Pentium FDIV bug, where certain floating point division operations gave wrong answers (I think they were slightly wrong).
If I know it makes mistakes I use it accordingly.That one was very famous, I remember. When I recall correct it was slightly wrong. But as you know, errors add up quickly in larger calculations.
This story is in deed very funny. And it shows how even errors can be unnoticed.
Interesting question was, why it stayed unnoticed.
All I recall is, some FPU/operations were not used, but people programmed their own, because the given ones provided not what was needed.
Now this is getting interesting. I wonder how my handsome version is in the universe where I'm handsome (if such a universe exists).because of the multiverse
Probably trolling poor OPs but with better avatar picture and STD induced tick.Now this is getting interesting. I wonder how my handsome version is in the universe where I'm handsome (if such a universe exists).
I've seen and had way more than enough of this "computers trying to anticipate/foresee/think for me helping me to don't make mistakes/know better than me, what I better going to do/suggest me some" - BS.Like UI - it is basically what LLM is - translator from human to computer.
rm -Rf /usr/* that has to be done - unquestioned.Now this is getting interesting. I wonder how my handsome version is in the universe where I'm handsome (if such a universe exists).
If I know it makes mistakes I use it accordingly.
Creators are fixing the models constantly. And it is getting better, much better. By model tuning methods and/or by the use of agentic processes i.e. self-testing feedback.
When you train a model very specifically i.e. on OS instructions - it will be more precise there.
However it seems the hallucinations etc. are inherent. Despite the hype, it is not a thinking machine, it is mimicking machine.
So, let's do it on the not critical stuff. Like UI - it is basically what LLM is - translator from human to computer. Fuzzy but good enough for most cases.
And that was the scope of muh original post - agentic UI.
When the LLM has to translate fuzzy human input in to code, it seems to me that it will make less mistakes on POSIX fully complaint system like FreeBSD than when trying to create meter long poweshell script just to read a processor type. On Linux it will make mistakes because of the multiverse of how distributions are glued together. And Mac will have no public training data.
You need to have seen this at least once bare-metal to pass an experience checkWe need a minimum age limit for computer use...I'm thinking 40?