In 2018 Gmail update, Google implemented several changes to the email platform, one of them is “Smart replies”, deep learning automated responses based on your email exchange style. The piece “a star which is not a planet is a sun not a planet” is based on Google automated email replies trained on the correspondence George Boole and Stanley Jevons sustained while Boole was developing Boolean logic. In these letters, Jevons proposed quite radical changes to Boole's system, which Boole did not accept, but later on become part of what we now understand as Boolean logic. Jevons can be seen as the first critic of Boolean logic since in a time it was published it didn’t receive much friction from others.

Machine learning algorithm is fed manually Booles and Jevons letters, and over course of time it learns and appropriates writing style of George Boole and offers quick replies to Jevons emails. The piece is a direct critique of simplicity of the “Smart replies” and the early history of mathematics that unknowingly influence machine learning of the second half of the 20th and beginning of 21st century. The title of the piece shows hidden beauty of mathematical thinking in the poetic and in the same time banal sentence “a star which is not a planet is a sun not a planet.”.

The title is appropriated from one of the letters sent by Jevon to Boole, where Jevon is disagreeing with the logical analysis of Boole:

“I do not think that either addition or subtraction is a process of logic, but that the operations of logic consist in combination and separation of terms or notions, by which all your inferences may be had. Thus your inference in p. 35 Stars except planets are suns, is really in logic the conclusion stars not planets are suns, or a star if not aplanet is a sun.

The true conclusion I venture to think is got by multiplying both sides by (1-z) or not planet. Thus

x=y+z
x(1- z) = (y + z)(l - z)
x-xz =y-yzfz-z
x-xz =y-yz

a star which is not a planet is a sun not a planet.”