Post

James Gleick
@around.com

The item you recommend is an example of the shoddy thinking to which I object. It is full of anthropomorphic language, accepting (without evidence) the idea that the "agents" are acting with intention and purpose. Show us the actual instructions and prompts provided by their human designers.

Gary Longsine 📱
@arbiteriapetus.bsky.social

The anthropomorphic language is the best fit we have to describe these systems. The trouble seems to be that (most?) people assume these expressions to be tightly coupled with moral responsibility, whereas programmers (generally) don’t.

James Gleick
@around.com

I don't understand why you think it's the "best fit." At best it's lazy; at worst misleading. Give me any example of using anthropomorphic language to describe actions taken by an AI model, and I will show you how it should have been written—without falsely implying agency, intention, or purpose.