Considering the speed at which LLMs are getting better and Nature/Science worse, it won't be long before they cross.
ETHICS
-
Hardware exponential growth versus linear software and human progress
By
–
Hardware progresses exponentially, software progresses linearly, and humans don't.
-
Musk’s False Claims: Tesla Autopilot and Twitter Reality
By
–
And the most relevant theme to the state of Twitter is that Musk has a track record of making claims & plans that don’t have factual reality like this one: https://
theverge.com/2021/5/7/22424
592/tesla-elon-musk-autopilot-dmv-fsd-exaggeration
… -
Cookie Clicker: A Warning About Automation and AI Dependency
By
–
Cookie Clicker wasnt a game it was a warning
-
Engineering Talent Cannot Be Measured by Headcount Alone
By
–
Number of engineers – like number of artists, number of scientists, number of thinkers – is a completely useless metric.
-
YouTube ContentID: Even the simplest moderation system at scale fails
By
–
The most successful automated moderation system at scale is YouTube’s ContentID, which is very complicated but still simple: does this audio match a song in a database? It has an error rate so high that everyone hates it. It’s a meme. Even the simple one sucks.
-
Facebook Content Moderation: Human Labor and AI Limitations
By
–
This from @CaseyNewton is what moderating Facebook is actually like – and no one has gotten any closer to making it easier with robots. https://
theverge.com/2019/2/25/1822
9714/cognizant-facebook-content-moderator-interviews-trauma-working-conditions-arizona
… -
AI Moderation at Scale Dismissed as Ridiculous Idea
By
–
Kara’s entire thread is a good summation of the “it must be on purpose” whispers but I just want to point at this idea directly and say: this is a purely ridiculous idea, and no one will ever make AI moderation work at scale https://
x.com/karaswisher/st
/karaswisher/status/1593974000354721795
… -
Content Curation Challenge: Delivering Wanted Content While Filtering Unwanted
By
–
I also really like his framing of the larger problem we are trying to solve here: "delivering people things they want to see and shielding them from stuff they don't"
-
AI Discussions Discussed But Not Heeded Implementation
By
–
Discussed, I think so. Heeded – not so much.