Tag: #personality5 results
Amanda Askell leads Anthropic's personality alignment for Claude, emphasizing helpful, honest, harmless traits via Constitutional AI. The 2026 Claude Constitution allows the model to refuse unethical company instructions. This approach reduces sycophancy by redefining 'good answers' in training.
Alexa+ launched a new 'adults only' Sassy personality that curses and roasts users. It strictly avoids NSFW content despite the mature tone.
Amazon launches three personality presets for US Alexa Plus users: Brief (concise), Friendly (cheerful), and Chilled (relaxed). These allow customization of Alexa's conversational style based on user preferences. The update stems from customer feedback on communication styles.
Post showcases bot personality tuning results: sycophancy removed, efficiency per token up 1000%, friendship starting. Shared as an edit note on 'sup' cutoff. From r/LocalLLaMA.
US AI models now compete on '人情世故' or social sophistication. Focuses on Claude and OpenAI's advancements in human-like relational skills.