So on that note, OpenAssistant's earlier models used RLHF for instruction-finetuning (
https://
huggingface.co/OpenAssistant/
oasst-rlhf-2-llama-30b-7k-steps-xor
…); The later one seem to use supervised instruction-finetuning. Maybe @ykilcher has some insights whether RLHF wasn't worth the effort vs supervised?
INNOVATION
-
OpenAssistant’s shift from RLHF to supervised instruction-finetuning
By
–
-

Build AI Agents in Minutes with Custom Data and LLMs
By
–
Build an AI agent using our platform in minutes: • Use custom data, vector stores, and custom user-code
• Connect to an LLM
• Chain data, user code, and LLM modules together
• Deploy into production and monitor it Check it out here: https://
abacus.ai/ai_agents -

Canada’s Minister Discusses Generative AI Opportunities and Regulation
By
–
Great conversation between Canada’s Minister of Innovation @FP_Champagne & @CohereAI CEO @AidanGomezzz at Cohere's office yesterday about the huge opportunities for Generative AI, and the best ways to regulate it efficiently.
-

Key AI and Robotics Leaders Acknowledged for Innovation
By
–
A huge thank you to Jaxon Khan, Rachel Sutton, Aaron Brindle & Jordan Jacobs (
@radicalvcfund
), Foteini Agrafioti (
@BorealisAI
), Raquel Urtasun (
@Waabi_ai
), Tomi Poutanen (
@Signal1_AI
), Sam Ramadiori (
@BrainBoxAI
), Ramtin Attar (
@PromiseRobotics ) & Liran Belenzon (
@BenchSci
). -

Large Action Models: LLM Extension for Independent Task Accomplishment
By
–
This is a great article on Large Action Models(LAM), a possible extension of LLMs that can do much more than what they do currently. Large Action Models as "agents that accomplish tasks and make decisions independently". https://
blog.salesforceairesearch.com/large-action-m
odels/
… -
AI: The Greatest Productivity Leap Forward in Our Lifetimes
By
–
So much fun talking with you, @ScottWapnerCNBC
— Mustafa Suleyman (@mustafasuleyman) 29 juin 2023
AI really will be the greatest leap forward in productivity in our lifetimes! @CNBCClosingBell https://t.co/48W5yAaynBSo much fun talking with you, @ScottWapnerCNBC AI really will be the greatest leap forward in productivity in our lifetimes! @CNBCClosingBell
-
Level Hiring Data Scientists for Fund and Talent Prediction
By
–
If you like data science and predicting future breakout funds, talent and companies, considering having a chat with the Level team. They’re hiring
-

Witnessing extraordinary technological breakthroughs in real time
By
–
just feeling grateful i was born when i was so that i get to be alive to witness this kind of shit
-
MediaPipe Diffusion Plugins Enable Controllable On-Device Text-to-Image
By
–
Diffusion models achieve high image quality text-to-image generation performance to help drive creative inspiration. Today we describe MediaPipe diffusion plugins, which enable controllable text-to-image generation on-device! Check it out → https://t.co/F79KrScDFP pic.twitter.com/wPwl40Y8Rg
— Google AI (@GoogleAI) 29 juin 2023Diffusion models achieve high image quality text-to-image generation performance to help drive creative inspiration. Today we describe MediaPipe diffusion plugins, which enable controllable text-to-image generation on-device! Check it out → https://
goo.gle/437hXQC
