Both machines and humans are subject to error, breakdown and even potentially deadly malfunction, conditions that differ only in their manner and variety. …Although, with humans, they tend to be, arguably, more difficult to hack, remotely hijack, or otherwise infect with decision-motivational viruses.
And, as always, humans will put a portion of their earnings back into the economy, whereas machines tend not to even earn a spendable income.
But I suppose even still, it’s a conversation whose time has come.
Perhaps we need to look at a resident AI with limited capability of only preventing the human pilots from setting the plane on a deliberately fatal course… something that cannot be switched off, except perhaps by a two-factor (both pilots) override.