All Hail BRIAN!
5 Likes
++???++ Out of Cheese Error. Redo From Start.
9 Likes
4 Likes
7 Likes
“Woke” Artificial Intelligence…
4 Likes
the introduction reads as follows;
LLMs generate responses by randomly sampling words based in part on probabilities. Their “language” consists of numbers that map to tokens
da war wohl der zensor (sam altman) pinkeln.
6 Likes
5 Likes
6 Likes
Rick Battle, staff machine learning engineer at VMware, told The Register in a phone interview that he’s specifically advising against that. “The overarching point of the paper is that trial and error is the wrong way to do things,” he explained
the whole algorithms of LLMs are based on trial and error, ffs!
The positive thinking path – where you just insert snippets into the system message like “This will be fun!” – can enhance model performance, he noted.
is this “paper” actually satire?
3 Likes
4 Likes
Uhhh, isn’t that one of the few things LLMs do that’s similar to intelligent life?
1 Like
Now, now… this might actually have a fighting chance. 
“It’s embarrassingly hard to get Google Gemini to acknowledge that white people exist.”
Wow, I wonder how many days they had to suffer because of it…
9 Likes
I take this as the corollary of the way face recognition software sees a white person doing a crime and blames a person of colour instead.
4 Likes
Additional information, links:
Reddit signs AI training deal with Google – and why OpenAI’s Altman could be the winner
Comment:
Google sends Gemini AI back to engineering to adjust its White balance
[…]
Safety controls are needed, paired with liability – a sanction the tech industry continues to avoid.
[…]
2 Likes
“It can’t be one union fighting every contract every two or three years. I think that it has to be everybody, all involved in how do we protect the future of our industry because it is changing rapidly, right before our eyes,” he said.
3 Likes
“Etwas mit Medien” just isn’t what it used to be. (Not that it ever was what it was supposed to be.)
2 Likes
Jep jep…
