[Funny] Even Clippy can be blamed on the use of non-Bayesian methods

From this 2001 ar­ti­cle:

Eric Horvitz… feels bad about [Microsoft Office’s Clippy]… many peo­ple re­gard the pa­per­clip as an­noy­ingly over-en­thu­si­as­tic, since it ap­pears with­out warn­ing and gets in the way.

To be fair, that is not Dr Horvitz’s fault. Origi­nally, he pro­grammed the pa­per­clip to use Bayesian de­ci­sion-mak­ing tech­niques both to de­ter­mine when to pop up, and to de­cide what ad­vice to offer...

The pa­per­clip’s prob­lem is that the al­gorithm… that de­ter­mined when it should ap­pear was deemed too cau­tious. To make the fea­ture more promi­nent, a cruder non-Bayesian al­gorithm was sub­sti­tuted in the fi­nal product, so the pa­per­clip would pop up more of­ten.

Ever since, Dr Horvitz has won­dered whether he should have fought harder to keep the origi­nal al­gorithm.

I, at least, found this amus­ing.