Episode
Greg Brockman: OpenAI and AGI
1:25:15
Published: Wed Apr 03 2019
Description
Greg Brockman is the Co-Founder and CTO of OpenAI, a research organization developing ideas in AI that lead eventually to a safe & friendly artificial general intelligence that benefits and empowers humanity. Video version is available on YouTube. If you would like to get more information about this podcast go to https://lexfridman.com/ai or connect with @lexfridman on Twitter, LinkedIn, Facebook, Medium, or YouTube where you can watch the video versions of these conversations.
Chapters
The co-founder and CTO of OpenAI discusses the intersection of AI and the physical world, and the goal of eventually creating safe and beneficial general AI.
00:00 - 02:23 (02:23)
Summary
The co-founder and CTO of OpenAI discusses the intersection of AI and the physical world, and the goal of eventually creating safe and beneficial general AI.
EpisodeGreg Brockman: OpenAI and AGI
PodcastLex Fridman Podcast
The speaker discusses the concept of psychohistory from Isaac Isamov's "Foundation," suggesting that the collective intelligence of society can predict behavior and the computer has been the most transformative innovation in that regard.
02:23 - 07:50 (05:27)
Summary
The speaker discusses the concept of psychohistory from Isaac Isamov's "Foundation," suggesting that the collective intelligence of society can predict behavior and the computer has been the most transformative innovation in that regard.
EpisodeGreg Brockman: OpenAI and AGI
PodcastLex Fridman Podcast
The creation of a powerful AI system entails understanding how to ensure its positive use and impact, and it is always easier to support the negative side of things than the positive side of things.
07:50 - 12:14 (04:23)
Summary
The creation of a powerful AI system entails understanding how to ensure its positive use and impact, and it is always easier to support the negative side of things than the positive side of things.
EpisodeGreg Brockman: OpenAI and AGI
PodcastLex Fridman Podcast
A lot of people stopped trying to dream about AGI technology due to the intractable nature of the technical safety problem, but researchers at OpenAI are pushing for innovative techniques to solve value alignment and empower humans more.
12:14 - 23:07 (10:53)
Summary
A lot of people stopped trying to dream about AGI technology due to the intractable nature of the technical safety problem, but researchers at OpenAI are pushing for innovative techniques to solve value alignment and empower humans more.
EpisodeGreg Brockman: OpenAI and AGI
PodcastLex Fridman Podcast
Aligning on resources, goals, and minimizing conflicts of interest are critical steps in building an AI company that can raise resources to achieve their objectives.
23:07 - 33:30 (10:22)
Summary
Aligning on resources, goals, and minimizing conflicts of interest are critical steps in building an AI company that can raise resources to achieve their objectives. It's important for companies like OpenAI to think beyond oneself, build a motivated team and get resources together to realize their most transformative technology yet.
EpisodeGreg Brockman: OpenAI and AGI
PodcastLex Fridman Podcast
OpenAI operates with a unique structure that allows for communication between employees and the CEO.
33:30 - 42:01 (08:31)
Summary
OpenAI operates with a unique structure that allows for communication between employees and the CEO. The power to decide the company's actions ultimately rests with the board.
EpisodeGreg Brockman: OpenAI and AGI
PodcastLex Fridman Podcast
The decision not to release the GPT-2 model has sparked conversations about the role of government in regulating AI and AGI development, from research to early and late-stage development.
42:01 - 47:42 (05:41)
Summary
The decision not to release the GPT-2 model has sparked conversations about the role of government in regulating AI and AGI development, from research to early and late-stage development.
EpisodeGreg Brockman: OpenAI and AGI
PodcastLex Fridman Podcast
OpenAI released a smaller version of GPT-2, sparking concerns over whether it is good for the world or not, due to people creating fake content, including Facebook messages and politician content.
47:42 - 52:31 (04:48)
Summary
OpenAI released a smaller version of GPT-2, sparking concerns over whether it is good for the world or not, due to people creating fake content, including Facebook messages and politician content. However, there are also creative applications, such as for sci-fi authors, and it serves as a test case for responsible disclosure of powerful models.
EpisodeGreg Brockman: OpenAI and AGI
PodcastLex Fridman Podcast
The idea of tying real-world identities to digital identities can help authenticate the source of content rather than the content itself, and GPT-2's understanding of the physical world is an interesting angle to explore in terms of natural language conversation and the Turing test.
52:31 - 1:01:09 (08:38)
Summary
The idea of tying real-world identities to digital identities can help authenticate the source of content rather than the content itself, and GPT-2's understanding of the physical world is an interesting angle to explore in terms of natural language conversation and the Turing test.
EpisodeGreg Brockman: OpenAI and AGI
PodcastLex Fridman Podcast
The speaker highlights the role of both large scale compute and small scale ideas in the progress of AI.
1:01:09 - 1:06:57 (05:48)
Summary
The speaker highlights the role of both large scale compute and small scale ideas in the progress of AI. While the large scale compute is geared towards pushing the limits of the technology and scaling up ideas, the small scale ideas often create breakthroughs and work better in smaller scales.
EpisodeGreg Brockman: OpenAI and AGI
PodcastLex Fridman Podcast
Dota serves as a step towards real-world problem-solving and demonstrates generalization out of distribution.
1:06:57 - 1:10:55 (03:57)
Summary
Dota serves as a step towards real-world problem-solving and demonstrates generalization out of distribution. However, caution should be taken as behaviors can emerge at a larger scale that were not present at small scale.
EpisodeGreg Brockman: OpenAI and AGI
PodcastLex Fridman Podcast
The speaker talks about how self-play has been able to push the professional level at the five versus five version of the game, running on a hundred thousand CPU cores, and they are still working towards the final competitive milestone for the project.
1:10:55 - 1:16:04 (05:09)
Summary
The speaker talks about how self-play has been able to push the professional level at the five versus five version of the game, running on a hundred thousand CPU cores, and they are still working towards the final competitive milestone for the project.
EpisodeGreg Brockman: OpenAI and AGI
PodcastLex Fridman Podcast
The speaker discusses the potential of adding individual projects and growing scale to achieve large teams of material engineering and ML science to create systems that work at large scale, ultimately leading to the creation of a simulator that can simulate the real world.
1:16:04 - 1:21:20 (05:15)
Summary
The speaker discusses the potential of adding individual projects and growing scale to achieve large teams of material engineering and ML science to create systems that work at large scale, ultimately leading to the creation of a simulator that can simulate the real world.
EpisodeGreg Brockman: OpenAI and AGI
PodcastLex Fridman Podcast
The speaker discusses the idea of consciousness in regards to AI and whether creating a conscious AI would help it survive and adapt in an environment in the same way as a living being.
1:21:20 - 1:25:02 (03:42)
Summary
The speaker discusses the idea of consciousness in regards to AI and whether creating a conscious AI would help it survive and adapt in an environment in the same way as a living being.