"We may be used to thinking of AI as tools, but some agents will be pursuing their own objectives. They will find ways to collaborate with people, by bargaining with, tricking or blackmailing them." 听起来就很AGI。
It's hard to believe that AI basically works by predicting the next word based on the input.
Today, AI can do many things remarkably well, such as writing code, answering questions, and creating articles.
What has happened in AI over the past few years to make it so capable?
Sometimes, I can solve it faster than the AI agent.
When I don't know how to fix a problem, I ask the AI agent to explain what is happening in my code and teach me how to solve it.
When I accepted all the code generated by an AI agent without reviewing it line by line, I started to lose controll of my project.
I have since found a better way to stay actively involved in the project. When I find a bug, I frist try to fix it by myself.
When I subscribe to a coding plan, I don’t really care how many tokens I use.
But when I use a pay-as-you-go API, I track every token and constantly check how much I’ve spent—even if it’s only one dollar.
Why does paying per use make me so much more sensitive to the cost?