Summary of Getting Started in AI Safety | Olivia Jimenez, Akash Wasil | EAGxVirtual 2022

00:00:00 - 00:40:00

This video covers the basics of getting started in AI safety, with a focus on how to get started. The presenter notes that if you're good at engineering, you can also test your fit for the field easily. They advise beginning AI safety researchers to follow their interests, read important literature, and engage in exercises to develop new, better ideas.

• 00:00:00 The speakers will discuss how to achieve goals in AI safety that are aligned with human interests, and warn of the potential existential risks posed by unaligned AI. They will also discuss traps new alignment researchers often fall into.
• 00:05:00 Olivia Jimenez discusses the importance of thinking about one's own "inside view" when it comes to AI safety, and how this can be aided by reading and keeping track of the literature on the topic. She also mentions Buck, Redwood's CTO, who has made similar claims about the need for people with a deep understanding of computer science and mathematics before attempting to address the challenge of AI alignment.
• 00:10:00 The video discusses the importance of starting problemsolving from the right angle, and how self-learning can help people do this more efficiently. It also discusses the importance of deferring to others when it comes to making decisions about what to do, and the dangers of being too eager to accept a project.
• 00:15:00 The video discusses the advantages and drawbacks of forward and back chaining when thinking about AI safety. Forward chaining is often ineffective, while back chaining can lead to more creative solutions. The presenter recommends practicing this exercise early in your AI safety journey, and using it when stuck.
• 00:20:00 Olivia Jimenez, Akash Wasil advise beginning AI safety researchers to follow their interests, read important literature, and engage in exercises to develop new, better ideas.
• 00:25:00 This video discusses different AI safety programs, and how to find the best one for you. It also mentions possible funding sources, such as grants, that may be useful.
• 00:30:00 The presenter advises beginners to dabble in AI safety research for a few months before committing to it fully, in order to build confidence and understanding of the field.
• 00:35:00 The interviewer discusses the importance of feedback in AI safety, and how following the "inside view" process can be a useful way of getting feedback. She also cautions that not everyone is suited for technical alignment work, and discusses some signals that can indicate if someone is not a good fit.
• 00:40:00 This video covers the basics of AI safety, with a focus on how to get started. The presenter notes that if you're good at engineering, you can also test your fit for the field easily.