People have been consistent about their opinion about level-5 autonomy. Five years back they said it was 5 years away, and they say the same thing today.
RESEARCH
-
Evaluating AI Alignment: Testing Jailbreaks on Most Advanced Models
By
–
in the past, I've tried to only post jailbreaks that work on the most advanced models on every question I can think of because to me that's the only fair assessment of current SOTA alignment methods and their limitations
-
Progress in Alignment Methods Despite Previous Reservations
By
–
coming from someone who clearly has lots of reservations about current alignment practices and has been very vocal about it, I actually think this example illustrates good progress in alignment methods here's why:
-
Uncanny Valley: Robotics, Freud, and Jasia’s Pioneering Translation
By
–
Super excited about this session on Wed: Jasia coined the term “Uncanny Valley” in her 1978 book by translating Masahiro Mori’s 1970 paper, identifying a subtle and important link between robotics and Sigmund Freud.
-
Snowflake’s Data Advantage for AI Model Fine-Tuning
By
–
For Snowflake, it gets a little interesting. They already host so much company data—the kind of data that’ll get used for fine-tuning—and just need a team that has experience with model dev/production to figure out that workflow.
-
Robotics Progress: From Leg Lab to Boston Dynamics Achievements
By
–
I was a PhD student in the Leg Lab. That was 40 years ago. Marc and @BostonDynamics have made remarkable progress but this compilation left out the bloopers and it misses @ylecun’s point that progress in LLMs does not necc. imply progress in robotics. https://t.co/dhvvCst1sd
— Ken Goldberg (@Ken_Goldberg) 26 mai 2023I was a PhD student in the Leg Lab. That was 40 years ago. Marc and @BostonDynamics have made remarkable progress but this compilation left out the bloopers and it misses @ylecun
’s point that progress in LLMs does not necc. imply progress in robotics. -
Barkour: Robotic Agility Benchmark for Quadruped Locomotion
By
–
Presenting Barkour, a robotic agility benchmark that tests the implementation of low-level locomotion skills useful for movement through real-world environments, along with an ML-based example generalist locomotion policy for quadruped robots. Learn more →… pic.twitter.com/DlXQiJx1Ne
— Google AI (@GoogleAI) 26 mai 2023Presenting Barkour, a robotic agility benchmark that tests the implementation of low-level locomotion skills useful for movement through real-world environments, along with an ML-based example generalist locomotion policy for quadruped robots. Learn more →
-
Single Universal LLM vs Multiple Specialized Models Strategy
By
–
Do you think there is going to be one LLM for all use cases or we will need to deploy multiple special (fine tuned) LLMs? @aesadde
-
Training LLMs with Code Execution Traces Improves Reasoning
By
–
People say that adding code to the LLM training corpus improves the model's "reasoning" ability. If so, would including a training set of short programs followed by an execution trace improve its reasoning further as it would have to predict the executed code path from the src?
-
LLM Panel Revival: Answering Audience Questions
By
–
We had such a great conversation about LLMs this week, but we weren’t able to get to all of our audience questions. We’re reviving the panel to let @pirroh @chipro @amasad answer some of them!