This argument seems flawed to me. Even if there are theoretical limits to how fully a computer program can analyze another, it does not immediately follow that this means a computer program is incapable of determining if another computer program is in fact a program running on a computer.

I don't need to fully understand all the ins and outs of the Linux source code to determine it's code, and that the code runs on a computer. I don't see why that would be different for a program.

Am I missing something?

Despite wanting to say "stopped clock wrong twice a day" I think he's right to conclusions. I might differ to reasons because my principle position is the lack of induction, and therefore inductive reasoning. Self recognition wasn't high on my list of reasons but I may reconsider.

Also, like lots of autodidacts out of his field, he overrated the Turing test. It's a thought experiment, its not real. We don't lock Chinese philosophers in caves Jared.