> Without reading the paper how the heck is agi mathematically impossible if humans are possible? Unless the paper is claiming humans are mathematically impossible?
Humans are provably impossible to accurately simulate using our current theoretical models which treat time as continuous. If we could prove that there's some resolution, or minimum time step, (like Planck Time) below which time does not matter and we update our models accordingly, then that might change*. For now time is continuous in every physical model we have, and thus digital computers are not able to accurately simulate the physical world using any of our models.
Right now we can't outright dismiss that there might be some special sauce to the physical world that digital computers with their finite state cannot represent.
* A theory of quantum gravitation would likely have to give an answer to that question, so hold out for that.
Finding something about physics that can't be perfectly represented is step one.
Then we also need evidence it can't be approximated to arbitrary quality.
And finally we need evidence that this physical effect is necessary for humans to think intelligently.