Robo-readers aren’t as good as human readers — they’re better

Annie Murphy Paul, The Thinking Stick, Aug 21, 2014
Commentary by Stephen Downes

When I read that robots are "unable to discern meaning" my first thought is to wonder what the critic thinks it is for a human to discern meaning. Yes, you can fool computers with nonsense - but you can also fool human referees of academic journals with nonsense as well. And - interestingly - it seems that it is becoming less and less easy to fool the computers, while humans remain as fallible as ever. So infallibility is not a criterion for being able to discern meaning.

This article suggests that computers may be better markers because they create a 'disinhibition effect' among students. "A non-judgmental computer may motivate students to try, to fail and to improve more than almost any human." But this isn't a criterion either - indeed, the author would not recommend allowing a computer to give grades. So what, then, is it to 'discern meaning' - and correspondingly, what is it to 'demonstrate meaning'.

I've discussed this in the past. Most writers believe that meaning (and truth) are based in representations, and that learning is essentially the creation (or construction) of these representations in the mind. So demonstration of meaning is a demonstration of the presentation and use of those representations. But this leaves the discernment criteria unfulfilled. Discerning is, I argue, a process of recognition. And computers can and do perform quite well at recognition tasks.

Views: 0 today, 122 total (since January 1, 2017).[Direct Link]