Counting AI by Hand ✍️ ~ 2. Agent Tokens
Mohsena, my PhD student, will lead a new lecture series on how to "count AI by hand" ✍️.
The first two lectures are:
Transformer Ops
Agent Tokens
Background
Mohsena's PhD research focuses on Explainable AI, with a particular interest in understanding and measuring AI cost, addressing questions such as:
How much memory does a transformer layer require?
How many tokens are taking up an agent's memory?
Where do the computational costs of AI actually come from
Approach
Instead of using equations and Greek symbols to calculate AI cost, Mohsena will show you how to represent operations and tokens as 2D rectangles in Excel. Once you can see them, you can measure them, compare them, and even count them by hand ✍️.
The goal is to make AI cost more intuitive, transparent, and understandable for everyone.
Sponsor
This lecture series is sponsored by By Hand Research.