src: click
Lisa. How can your selfimage be so much that of a journalist - yet you suck so much at the intelligence thing. How?
Well - Die Presse doesnt mind, and pays her a full salary. Thats a win for Diversity.
edit: Für die Idioten unter meinen Lesern:
“I want my matrix multiplication math done by a women!”
“I want my reinforcement learning gigworkers in india at least half female, and half Dalits!”
“I want initial prompts to be engineered, more like the AMS engineered them!” see: click
“I want at least 51% of the training data to be diverse literature.”
“I want diverse people to write at least an additional half of whats in all literature, videos produced, magazines printed, and on the internet, but they have to be qualified, so I want to test them first. No jokers apply. Oh, and I need at least 100 diverse van Gogh types.”
“Also, I need a PR person, to fix the public perception of this: click” (Messing with weights.)
Please for the sake of the argument, someone fetch all feminist and diverse classified literature, and then make sure its 60% of the training data, then fill up the rest with general knowledge writings, and spin up a foundation model for an LLM. Train it with Deepseek R1 671b q4 as a judge. Make sure you call it “Diverse Lisa”, and mail api access to Lisa Schöttel.
Can you do me that favor?
I mean -- this learning opportunity would benefit everyone, not just the diverse community - so it ought to be much easier to generate the goodwill (and funding) around it so some one spins it up, right?
(Its not as if you had to convince an angel investor for that… Hardware cluster access for a week maybe…)
edit: Several niche expert LLMs, one of which is a diversity expert. Then pull them together as a “general” model. Problem. The bubble is not tapped into for most requests.
edit2: Meso level issue: Diversity experts told the entirety of society to tap into Diversity for a 3% productivity increase. So they sent away the STEM Cracks from china - from western elite universities to make a more diverse quota.
Now china just increased LLM efficiency (gaged by price, because o3 is not open source) by 27x. Roughly. And made the model more or less open source. (Not the training data.)
Now every KI company (that invested in NVidia Cuda cores) in the world is like: FFFFFFUUUUUUUUUUUUUUUUUUUUUUUCCCCK the diversity team. (As in its not just data and compute you needed. Algo development was up there as well.)
Especially their angel investors.
So is every social media company in the world.
There is a fun little overlap.
edit: Better discussion about the topic:
