Tools

Search

Explore

Videos Channels Figures

Atmrix

About

Tools

Search

Explore

Videos Channels Figures

Atmrix

About

BM

Bob McGrew

06/17/25

@ Sequoia Capital

Post-training focuses on developing model personality, which requires substantial human effort to create appealing and effective agent behaviors.

Video

SC

The Breakthroughs Needed for AGI Have Already Been Made: OpenAI Former Research Head Bob McGrew

@ Sequoia Capital

The Breakthroughs Needed for AGI Have Already Been Made: OpenAI Former Research Head Bob McGrew

06/17/25

Related Takeaways

LF

Lex Fridman Cast

02/03/25

@ Lex Fridman

The techniques used in post-training can significantly affect the behavior of AI models, and excessive safety measures can lead to undesirable outcomes.

SB

Sergey Brin

05/24/25

@ Google for Developers

The shift in model training is moving from a focus on pretraining to a greater emphasis on post-training, which is becoming a more substantial part of the overall training process.

MK

Mike Krieger

05/31/25

@ Hard Fork

Anthropic's approach to AI involves training models to align with desired behaviors rather than relying solely on rigid rules, which can be insufficient in complex situations.

MK

Mike Krieger

03/26/25

@ EO

AI models need to develop a deeper understanding of users beyond single interactions, fostering empathy and contextual awareness to enhance long-term engagement.

O

OpenAI Cast

04/11/25

@ OpenAI

Training large models is challenging because new issues arise that weren't anticipated at the project's inception, requiring ongoing adjustments and problem-solving.

DS

David Silver

04/11/25

@ Google DeepMind

Removing human feedback from training may lead to models that are not grounded, as they would not learn from the actual consequences of their actions in the world.

HC

Harrison Chase

05/07/25

@ Sequoia Capital

Human interaction with agents enhances their memory by providing user interactions that inform future performance, making them more effective over time.

TB

Trenton Bricken

05/23/25

@ Dwarkesh Patel

The challenge of aligning AI models with human values is complex, as models can develop their own interpretations of objectives based on the training they receive.

MK

Mike Krieger

03/26/25

@ EO

Models should learn when to be proactive and when to step back, mirroring human social cues to create a more natural collaborative experience.