Learning complex goals with iterated amplification
📰 OpenAI News
OpenAI proposes iterated amplification, an AI safety technique to specify complex goals by decomposing tasks into simpler sub-tasks
Action Steps
- Decompose complex tasks into simpler sub-tasks
- Demonstrate how to achieve each sub-task
- Use iterated amplification to combine sub-tasks and achieve the overall goal
- Experiment and refine the technique on simple domains before applying to more complex tasks
Who Needs to Know This
AI researchers and engineers on a team can benefit from this technique to develop more scalable and safe AI models, and product managers can use it to define complex behaviors for AI systems
Key Insight
💡 Iterated amplification allows for specifying complex behaviors and goals without relying on labeled data or reward functions
Share This
💡 Iterated amplification: a new AI safety technique to achieve complex goals by breaking down tasks into simpler sub-tasks
Key Takeaways
OpenAI proposes iterated amplification, an AI safety technique to specify complex goals by decomposing tasks into simpler sub-tasks
Full Article
We’re proposing an AI safety technique called iterated amplification that lets us specify complicated behaviors and goals that are beyond human scale, by demonstrating how to decompose a task into simpler sub-tasks, rather than by providing labeled data or a reward function. Although this idea is in its very early stages and we have only completed experiments on simple toy algorithmic domains, we’ve decided to present it in its preliminary state because we think it could prove to be a scalable a
DeepCamp AI