shmval

5 ms·

1.
The Universe just wants to learn(dayafter.substack.com)
5 points by shmval 3mo ago|0 comments
2.
by shmval 4mo ago
I'm running these AI visibility analyses across different niches, the report goes into the methodology details. AI coding tools shows the disparity between real on-the-ground sentiment and what AI responses suggest much more glaringly
3.
Which AI coding tools do AIs recommend? Data across 10 models and 1000 responses(tryrenown.com)
2 points by shmval 4mo ago|1 comments
4.
by shmval 4mo ago
The message here obviously comes from a good place. But it can't escape the trap baked into its own theology. Putting humanity one rung below God, with everything else below us is a trick that allowed semitic civilizations to make empi
5.
Into the Unwritten Dawn(dayafter.substack.com)
2 points by shmval 1y ago|0 comments
6.
by shmval 1y ago
The trouble is there is a clear dichotomy in the scenarios where the LLM pushes back like this, most likely derived from RLHF. So anything health related, and "touchy" subjects in general will face this. But random design choices
7.
by shmval 1y ago
Yes but you have to want that, and most people do not. Therein lies the rub.
8.
by shmval 1y ago
This. I think it's the key.
9.
by shmval 1y ago
I think there's a lot of alpha left in building a better and more intuitive UX for seed/top-p/temperature etc. The vast majority of users don't get that far.
10.
by shmval 1y ago
Better prompting does provide more balanced responses to a certain extent but users looking for validation often subconsciously leave bread crumbs that the more powerful models pick up on.
11.
The Emperor's New LLM(dayafter.substack.com)
94 points by shmval 1y ago|62 comments
12.
Thoughtsourcing(dayafter.substack.com)
2 points by shmval 1y ago|1 comments
13.
Multilingualism and Cognitive Advantages for LLMs(dayafter.substack.com)
1 points by shmval 3y ago|0 comments
14.
Infrastructure troubleshooting powered by machine learning(youtube.com)
2 points by shmval 4y ago|0 comments