scripod.com
OpenAI Co-Founder: AI Goes Parabolic! Here's What's Next | Greg Brockman

Overview

Shownote

Highlights

Transcript

Chapters

Pins

OpenAI Co-Founder: AI Goes Parabolic! Here's What's Next | Greg Brockman

The Knowledge Project

5 DAYS AGO
The Knowledge Project

The Knowledge Project

5 DAYS AGO
OverviewShownoteHighlightsTranscriptChaptersPins

Shownote

The AI race, the future of AGI, and the inside story of OpenAI. Greg Brockman is the co-founder of OpenAI. This is the most detailed first-person account he has given of the 72 hours after Sam Altman was fired, how OpenAI started, and the future. Greg e...

Highlights

In this candid and revealing conversation, Greg Brockman—OpenAI co-founder and former president—offers a first-person account of the organization’s founding, its pivotal technical and structural decisions, and the dramatic leadership crisis following Sam Altman’s firing.
00:00
Making a difference in AI would make for a well-lived life
00:49
Sam Altman realized the speaker had already decided to start an AI company
02:40
The Napa off-site produced the technical plan that has guided OpenAI for 10 years
04:25
Google DeepMind seemed to have an insurmountable advantage in the field before AlphaGo's release
04:54
Creating a for-profit entity was the only path to achieve OpenAI's AGI mission
06:05
Machines could learn semantics
08:22
By scaling PPO, they exceeded the performance of the best humans, showing that massive compute with simple algorithms works in practice
10:05
Reasoning and prediction are deeply connected to intelligence
12:00
Conflicts within the AI field take on existential weight due to the high-stakes nature of OpenAI's mission
15:44
They were removed from the board immediately after learning Sam would be removed
17:50
On Sunday night, the board replaced the interim CEO, causing the company to rebel and creating chaos
19:56
Many people canceled their Thanksgiving flights and gathered at the office to be part of the historical moment
23:18
No one left for more money or better offers despite poaching attempts
23:49
Trained language models on DNA sequences to advance health research
28:03
It's painful but worthwhile to create an environment where others can do great work
28:23
Suffering is necessary to build value
32:23
AI is being applied to its own development process, making it faster
33:26
AI is now much better than humans at writing code, though humans still excel at structuring code
36:21
Technological improvements ensure AI is aligned with users' long-term goals rather than short-term satisfaction
38:08
We're in a global AI renaissance
38:40
Leading in AI is critical for the US to protect democratic values
39:49
The core advantage lies in the 'machine' that creates models, not just individual models
40:39
Training the model to have a good-looking chain of thought may lead to loss of faithfulness
41:47
The current trend is to release preview models due to a compute-constrained world
43:38
OpenAI's core is to face reality and think about long-term implications
46:32
Dedicated data centers for specific problems like cancer research could happen this year
47:52
Empowering people with the technology is core to our mission, rather than focusing solely on solving problems in an ivory tower approach
51:32
Everyone on the planet should have access to a personal AGI that knows them well, can provide advice, and help achieve goals
59:04
Safety involves long-term thinking, including model training and feedback loops, and extends to building societal resilience for AI
1:03:59
Data centers use little water due to a closed-loop system
1:04:40
We need to solve certain problems and ensure people can feel the impact in their daily lives.
1:04:45
AI is about empowerment and human agency, not just automation
1:07:17
Leaning into AI technology will be a critical skill in the future, allowing people to manage AI agents and achieve goals more easily
1:11:45
Success means achieving OpenAI's mission of making AGI benefit all of humanity

Chapters

Introduction
00:00
Meeting Sam Altman and Starting OpenAI
00:49
Building the Founding Team
02:40
DeepMind's Lead Over OpenAI
04:25
The Change from a Pure Non-Profit
04:54
Breakthrough Moments at OpenAI
06:05
What Dota 2 Meant for OpenAI
08:22
Reasoning Versus Prediction
10:04
Tensions Grow at OpenAI
11:59
Sam Altman's Firing
15:44
Greg Quits OpenAI
17:49
Sam Explores Deal with Microsoft's Satya
19:56
OpenAI Employees Sign Petition for Altman's Return
20:28
Ilya Sutskever Leaves OpenAI
23:43
Lessons Learned in Leadership after Sam Ousting
24:59
The Thing Ilya Said that Greg Can't Forget
28:22
Is AI Going Parabolic?
32:22
How Much of OpenAI's Code is Written by AI?
33:24
Are AI Chatbots Just Telling Us What We Want to Hear?
36:21
The Global AI Race to Reach AGI
38:06
What Happens if US Doesn't Reach AGI First?
38:40
Are Competing Countries Stealing AI Advancements from U.S?
39:49
Why ChatGPT No Longer Shows Reasoning
40:38
The Finite Constraints of Compute
41:47
On Investing Early in Data Centers
43:38
The Future of Data Center Specialization
46:31
How OpenAI Will Decide Whose Queries to Serve
47:52
OpenAI on Consumer vs Enterprise Models
49:08
Data Centers in Space?
53:05
What Should AI Regulation Look Like?
1:00:56
The Future of AI-Powered Entrepreneurship
1:04:33
AI and Job Loss
1:04:44
The Skills Young People Should Invest In
1:07:15
What Does Success Look Like For You?
1:11:30

Transcript

Shane Parrish: So how did OpenAI come about? Greg Brockman: I knew I wanted to do a startup because I felt like that was something, But you were just in a startup. Stripe was a startup, It's true, but I felt like Stripe. The problem that we were solving, ...