What Lambda Is and How Does It Think?

Lambda can make a call to YouTube. The language that it uses to talk to YouTube, it's not plain English. So the question is, where do you draw the boundaries of what Lambda is? Do you just draw it around the language model? Or do you actually include the vision processing systems that it's talking to in YouTube? And depending on where you draw those boundaries, you have different models of what LambDA is and how it's thinking.