To be fair, the point Gwern's making (not shown in the quote) isn't just "I don't like it" — it's that GPT-3 is capable of imitating pathological preferences in the reward model.
AI
-
AI Breakthroughs 2022 Deep Learning Year in Review
By
–
¡1er VÍDEO DEL AÑO! 2022 : EL AÑO DE LA INTELIGENCIA ARTIFICIAL Pero… ¿Por qué? Hoy os traigo el resumen perfecto que recoge los avances más interesantes del mundo del Deep Learning del año pasado LINK: https://
youtu.be/rKJdwLkPPPM Bienvenidos a mi AI-REWIND 2022 -
Follow DAIR AI for Trending ML Papers Weekly
By
–
Follow us @dair_ai for upcoming trending ML papers of the week.
-
A Succinct Summary of Reinforcement Learning
By
–
A Succinct Summary of Reinforcement Learning. 11 of 11
-
Integrated Decomposition Improves Science Q&A with Human-in-Loop
By
–
Proposes integrated decomposition, an approach to improve Science Q&A through a human-in-the-loop workflow for refining compositional LM programs. 10 of 11
-
Deep Learning Memorization: Understanding Overfitting Phenomena
By
–
This work aims to better understand how deep learning models overfit or memorize examples; interesting phenomena observed; important work toward a mechanistic theory of memorization. 8 of 11
-
StitchNet: Novel Neural Network Paradigm Through Pretrained Fragment Reuse
By
–
StitchNet is a novel paradigm to create new coherent neural networks by reusing pretrained fragments of existing NNs. 9 of 11
-
ConvNeXt V2: Performant CNN Model with Masked Autoencoder Framework
By
–
ConvNeXt V2 is a performant model based on a fully convolutional masked autoencoder framework and other architectural improvements. CNNs are sticking back! 6 of 11
-
Large Language Models Used for Corporate Lobbying Activities
By
–
With more capabilities, we are starting to see a wider range of applications with LLMs. This paper utilized large language models for conducting corporate lobbying activities. 7 of 11
-
Building Useful Tools for Builders and Community
By
–
We're hard at work to build something that we think is useful to builders & the community and we'll see what happens! Never hesitate to tell us if you have ideas of things we should build/do to improve, it's very welcomed!