Audio should be easier. You could for example only transcribe voices that match a voice that has given permission in that voice. You’d still need AI to analyze whether it was AI generated but I suspect doing that on voice is easier than text.
RESEARCH
-
Awesome Post on Devising ML Metrics by Dan
By
–
Just read, this is an awesome post Dan! (Originally thought it was just an image lol, here is the post: https://
safe.ai/blog/devising-
ml-metrics
…) -
From Simpson Era to Futurama Era of AI Predictions
By
–
we are moving from the “Simpson era” of cartoon predictions to the “Futurama era”
-
LLM Bugs: Finding and Discussing Detection Methods
By
–
LLM bugs and how to find them Part of the joy of curating @aidotengineer is I get to bring some of the best people in the community that you only see online, to meet and teach in person. Several of our speakers are giving their first talks -ever-! Daniel is one of the most
-
LLMs trained on specific Myers Briggs personality type text
By
–
Random idea: LLMs trained on written words only from people of a specific Myers Briggs type.
-
LLMs Understanding World vs N-gram Models: Insufficient Argument
By
–
"LLMs understand the world because that's needed to predict the next word" is an insufficient argument: you could say the same about n-gram models.
-
Open Access Publishing: Balancing arXiv Visibility with Peer Review
By
–
And if you have to do one more click to get to the open arXiv version, that hardly seems like a hardship. OTOH, people often just miss the peer reviewed version because they get sent directly to arXiv.
-
Design choices for peer-reviewed research accessibility and transparency
By
–
There are many many design choices that could help readers differentiate between peer-reviewed and non-peer-reviewed work while also pointing to accessible options when the peer reviewed version is paywalled.
-
ArXiv Citations Prioritized Over Peer-Reviewed Publications in AI Research
By
–
AND they point to arXiv first even if the paper has been published in a peer-reviewed venue.
-
Programmatic Data Annotation for Model Alignment and Curation
By
–
Instead of manually doing thumbs up and thumbs down, we can programmatically annotate what makes something acceptable vs. not, and that data can be used to align a model or curate data for fine-tuning. More details in the Snorkel Blog: