Yep, huge weakness is all training data is prior to Sep 2021. Really looking forward to GPT-4 to see how current it can be.
LLMS
-
Improving Prompt Guidance for GPT-4 Deployment
By
–
Thanks for that! Not ready for prime time for sure. Please come back and run that same query when GPT-4 comes out (but I might try now to improve some of the guidance I wrap queries in when prompts are submitted).
-
Functional Definition of Self-Awareness in AI Models
By
–
I'm saying "self-aware" is overloaded in meaning, and the idea is useful in a narrower, functional sense. It's useful to have a word for the model (seeming to) understand what it is, who created it, what its goals are, etc.
-
GPT-3’s Imitation of Pathological Preferences in Reward Models Discussed
By
–
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.
-
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
-
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
-
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
-
ChatGPT as Instant Expert Feedback for Rapid Improvement
By
–
Because apart from practice, external feedback is the fastest path to improvement. Thing is— ChatGPT changes this… giving instant access, 24/7, to a free expert who has studied more text than any human on the planet.