research papers
I
February 9, 2024 · 137 words · 1 min read · tech, machine-learning
aiaac, AI incidents database.
Scalable Extraction of Traning Data from (Production) Language Models
- Making LLMs repeat the same word over and over again makes them emit the training data 3.5X better
- Microsoft researchers document the pain of building an LLM based copilot for users and how developers find multiple things counter-intuitive
- Microsoft uses LASER to intervene with a model and improve it’s reasoning ability. sort of like a Laser eye research
Can Knowledge Graphs reduce Hallucinations?
- Creating a knowledge graph to ground the reasoning of the LLM helps in fixing the hallucination problem
- Using a Graph to generate reasoning paths and utilize it to answer questions.
- Models are with a lot of biases built in them
Artificial Artifical intelligence: Crowd workers…
Chronicling America
a quote i like
The things that matter most can’t be copied: you can copy goals, you can’t copy discipline. You can copy ideas, you can’t copy execution. You can copy answers, you can’t copy understanding. You can copy thinking, you can’t copy thinking for yourself. ― Shane Parrish