Partnering with Constitutional AI
I partner with constitutional AI every day. One of my go-to tools is Poe.com, where I can use multiple LLMs as brainstorming partners, all in one interface.
- Ask Sage's Sage
- Meta's Llama 2
- Google's PaLM 2
- OpenAI's GPT-3.5/4
- Anthropic's Claude 2 (constitutional)
Here's what I do: I thoughtfully write a prompt and pop it into each bot to see what different responses I get. It's like having smart colleagues with unique views on a topic, all hanging out in my office waiting to help me.
What Makes Claude Stand Out?
What is Claude's standout feature that has me using it daily? A whopping 100K tokens per prompt. That's 12 times the standard with capacity for 75,000+ words, roughly the size of a business textbook. This allows it to understand more context and produce more diverse and coherent outputs. Examples of how you could use it:
- Research
- Industry Analysis
- Customized Solution Creation
- Contract Review
- Competitive Analysis
- Customer Feedback Analysis
- Training and Onboarding
- Strategy Collaboration
- Monitoring Regulatory Compliance
- Knowledge Store Accessible by Natural Language Questions
How Claude Is Different
In addition to using humans to fine-tune Claude like other LLMs do, Anthropic also uses a second AI model called Constitutional AI. Intended to discourage toxic, biased, or unethical answers and maximize positive impact, it includes:
- Rules borrowed from the United Nations' Declaration of Human Rights and Apple's terms of service.
- Rules that Claude's researchers found improved the safety of Claude's output, like "Choose the response that would be most unobjectionable if shared with children."