The cornerstone of philosophy is the ability to communicate ideas clearly and precisely.
It’s also the key to getting the most out of an AI model, according to Anthropic’s resident philosopher.
Amanda Askell, a member of Anthropic’s technical team and a trained philosopher, says effective incentivizing requires finding the right balance between multiple considerations.
On Anthropic’s “Ask Me Anything” podcast, Askell, who studied philosophy at Oxford and New York University, according to his LinkedIn, explained his thought process.
“It’s really hard to understand what’s going on because one thing is like the willingness to interact a lot with the models and really look at production after production,” she said.
Good blowers should be “very experimental,” she said.
However, the incentive goes beyond experimentation, and that’s where his philosophical training came in handy.
“That’s where I actually think philosophy can be helpful in incentivizing in some way, because, for example, a lot of my work is just trying to explain an issue or a concern or a thought that I need to present to the model as clearly as possible,” she said.
This emphasis on clarity is important not only to help people refine their own prompts, but also to understand the AI itself.
In a “Prompt Engineering Overview” released by Anthropic in July, the company said users interacting with Claude, its chatbot, should view him as “a bright but very new employee (amnesiac) who needs explicit instructions.”
“Claude has no context about your standards, styles, guidelines, or preferred ways of working. The more specifically you explain what you want, the better Claude’s response will be,” Anthropic wrote.
Venture capitalist veteran Marc Andreessen said last month that the power of AI comes from treating it like a “thought partner.”
“Part of the art of AI is what questions to ask it,” he said.
Those who master this skill can land lucrative jobs as speed engineers, which have a median salary of $150,000, according to levels.fyi, a platform for tech workers to research and compare salaries.
