This is what happens when you release a poorly tested #LLM in the hands of real people who now suffer the consequences @OfficeforAI @SciTechgovuk @theCAIDP @GPAI_PMIA
LLMS
-
Rodney Brooks Shares Thoughts on GPT Models and Transformers
By
–
Lots of people have been asking for my thoughts on Generative Pre-trained Transformer (GPT) models. I have now written them down. https://
rodneybrooks.com/what-will-tran
sformers-transform/
… -
RLHF’s Impact on Understanding Pre-trained LLM Behavior
By
–
RLHF has gotten so good we no longer develop an intuition about how pre-trained LLMs behave. To the surviving flashes of old behavior we wrongly ascribe meaning — humanity, madness.
-
Microsoft UPRISE Boosts LLM Zero-Shot Performance via Prompt Retrieval
By
–
Microsoft’s UPRISE Automatically Retrieves Prompts to Boost the Zero-Shot Performance of Large Language Models https://
syncedreview.com/2023/03/22/mic
rosofts-uprise-automatically-retrieves-prompts-to-boost-the-zero-shot-performance-of-large-language-models/
… -
GPT-4 access limitations frustrate users with restricted message quotas
By
–
I would ask GPT4 for you, but I only got 25 precious messages every 3 hours
-
ResNet and GPT-3 achieve efficiency gains matching larger models
By
–
Some of the highlights of this work include ResNet-18 on ImageNet achieving a 3.5% accuracy improvement, and #GPT-3 Small on WikiText-103 reducing perplexity by 0.4, both matching larger dense model variants that have 2x or more FLOPs. (3/4)
-
Important Preprint on LLMs and Data Rivers
By
–
Important preprint from Sylvie Delacroix on LLMs and Data Rivers. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4388928
… -
GPT-4 vs ChatGPT: Users Share Experiences with Advanced AI Models
By
–
Vous êtes passé à #GPT4 ou pour ne pas payer 20 dollars par mois vous êtes resté à #ChatGPT ? Personnellement je suis passé à #GPT4 et je suis ÉPOUSTOUFLÉ C’est vraiment révolutionnaire
-
Prompt Engineering for Rapid LLM-Based Product Prototyping
By
–
Prompt engineering for rapid iteration in product prototyping and exploration tailors a LLM to better align with the task definition quickly and easily. @lucasmearian reports via @Computerworld with quote from @MarshallChoy
. https://
computerworld.com/article/369125
3/how-to-train-your-chatbot-through-prompt-engineering.html
… #ai -
Bard’s Training Data Transparency Issues and Speculation
By
–
Bard still seems confused about whether it was or wasn't trained using private data from Gmail (Google says it it wasn't). In reality it probably has zero idea what it was trained since that information was not in the training data, so it's just making guesses.