You are a project manager, leading the AI alignment project at Aurora Media, where your team is refining a brand-language model to generate emotionally engaging ad copy. Your data scientists have collected human preference labels comparing which ad responses sound more authentic.They are now deciding whether to apply Reinforcement Learning from Human Feedback (RLHF) or Direct Preference Optimization (DPO) for alignment. Which step is specific to RLHF and not part of DPO’s training procedure?

As a product manager, you want your sales-assistant LLM to sound slightly more enthusiastic, but your team has limited computing resources and needs results quickly. You already have a trained Sparse Autoencoder (SAE) that identifies interpretable tone features.  What should you do next?

As a product manager, you must explain Direct Preference Optimization (DPO) to executives. Which analogy fits best?

A global luxury brand discovers that its marketing emails sound “too casual” for customers in Japan and the Middle East, where a more formal tone signals respect and professionalism. As the brand manager, how would you adapt the brand voice across regions when you have limited computation budgets? Choose the best option among the following considering performance and computation cost. 

What did the investment bankers do during the postbellum era?

Why would large companies want to merge vertically with other companies? Give 2 reasons.

What did those not in favor of Taylor’s system say?

What is the advantage of a moving assembly line?

What is antitrust?

Why would small companies want to merge horizontally with each other?