When it comes to the blending of human and AI interaction, it's essential to remember that at the end of the day, the AI has limitations based on the inputs that we choose to expose it to, and we decide what it's allowed to say and do. It's crucial to consider what kind of future we want when designing AI-human interaction and to question what the objective function should be.
Humans are the dominant builders of information, but we are now creating increasingly sophisticated forms of AI that are shaping our future. It is important to understand the characteristics, limitations, and potential improvements of these systems as they continue to evolve.
The speaker believes that institutions should prevent corporations from rigging the system, so they act in the interest of the people and protect the environment. The investment in AI safety research should be accelerated without slowing down technology.
OpenAI is working on technical safety mechanisms to ensure that systems built align with human values, by looking at data and building proof of concept models that can learn human preferences.
The speaker discusses the importance of defining core values and sticking to them when making decisions about who to hire and which industries to engage with.