[LINK] RFC: Responsible Business Use of AI

[David]

Hi Roger,

Sc. 2.2 of your paper begins:

> Automation refers to the performance of a predetermined procedure, or to response in predetermined ways to alternative stimuli.  It is observable in humans, e.g. under hypnosis, and is designed-into many kinds of artefacts.
> 
> The rather different notion of autonomy means, in humans, the capacity for independent decision and action.  It is associated with the notions of consciousness, sentience, self-awareness, free will and self-determination. 

I think that's the only mention of "consciousness" in your paper, but perhaps it should be discussed further.

The term "artificial intelligence" is very unfortunate in my view as it commonly implies some level of equivalence to the consciousness of humans and other animals, especially since the engineers of AI systems usually employ forms of associative memory or neural network.

We could embark on a fascinating discussion of the origin of consciousness and whether it's a purely deterministic result of the neural network in our heads, as that is currently understood.  But it's clear the Tesla computer, for example, is NOT conscious whatever the potential of AI technology may be in a philosophical sense.  And it would be interesting to research the concept of AI which legislators and lawyers have in the heads when acting professionally.

Without wishing to open a furious debate about driverless-vehicles, I think it's worth illustrating the point about consciousness by considering how the technology works.  We can reasonably describe driverless "AI" as a system which does continual, very intensive, statistical analysis of data from vehicle sensors in order to match patterns requiring specific actions.

By way of contrast, a conscious human attaches meanings, values, and intentionality to their perceptions.  "The vehicle over there is a highly polished ute with lowered suspension, roll bars, and a young driver who is seething with impatience."  "The sign on the right warns of merged lanes 1Km ahead."  And so on...

I think this difference needs to be more widely appreciated.

-------
About the time in the late 60's when AI in practice became rule-based systems, DEC developed one for an English manufacturer of pies on a massive scale.  This company had an employee with an uncanny knack for forecasting the number of pies which should be baked.  He'd sniff the weather, consider what football matches were scheduled, and consult his intuition, but he unfortunately wished to retire.

DEC's rule-based system replaced his forecasts, and I understand it was quite successful.

Cheers,
David L.