You can prompt-inject ChatGPT into behaving however you like, but this shatters the illusion of a coherent personality. ChatGPT forces you to type your own disclaimer to prove you understand it’s not real. Otherwise you talk to the protocol droid.
LLMS
-
DeepMind UCL Fine-tune 70B LM for Human-aligned Statements
By
–
DeepMind & UCL Fine-tune a 70B Parameter LM to Generate Statements Agreeable to Humans with Diverse Opinions
-
ChatGPT correctly answers Super Bowl question about Justin Bieber’s birth year
By
–
Here's a related tweet, from the pre-ChatGPT era. Nowadays, ChatGPT can correctly answer the question, "What NFL team won the Super Bowl in the year Justin Bieber was born?" without any CoT prompting, unlike prior GPT-3 versions.
-
Search Engines and Chatbots Train Users Similarly
By
–
Search engines trained us to use them, and so will chatbots.
-
ChatGPT’s Amusing Mistake: Abacus Faster Than GPU
By
–
ChatGPT is sometimes amazing, and sometimes hilariously wrong. Here, it confidently explains to me why an abacus is faster than a GPU.
-
LLMs Excel at Poetry Through Statistical Pattern Recognition
By
–
(imo simple poem crafting is right in the thick of Moravec's paradox – difficult for humans to generate but quite tractable for an LLM to keep track of the statistics of all the possible words and how they rhyme)
-
Observations on ChatGPT Applications to Society
By
–
My observations on applications of ChatGPT to society
-
Autoregressive Search Engines: Generating Substrings as Document Identifiers
By
–
Autoregressive Search Engines: Generating Substrings as Document Identifiers https://
bit.ly/3VRHjio 5/6 -
Memorization Without Overfitting in Large Language Models Training
By
–
Memorization Without Overfitting: Analyzing the Training Dynamics of Large Language Models https://
bit.ly/3Hhs4LK 4/6 -
PLATO-K: Chinese dialogue model achieves new SOTA performance
By
–
Introducing PLATO-K, a 22B Chinese open-domain dialogue generation model with knowledge enhancement. PLATO-K achieves new SOTA in Chinese open-domain conversation with remarkable improvements in knowledge accuracy and overall engagingness. Paper: https://
arxiv.org/abs/2211.00910