Was referred this by a friend who knows well my mania for 'The Definiton Problem' in AI.
My experience maps onto yours pretty directly in respect of the definitional confusions I encounter within applied AI - which naturally get worse the further out you go from a centre filled with engineers on more intimate terms with the tech - but I think this is also a much wider issue in theoretical AI too.
Just as there's a lack of terminological definition in ML vs. DL, to use your for-instance, there's a lack of terminological consensus in AI research about what, for instance 'intelligence' (/'learning'/'consciousness' etc.) actually means. In most respects, participants in these sorts of conversations don't even realise that their definitions a. differ and b. are not at all rigorous in and of themselves. Shy of such essentials, the nearness-at-hand of the pathbreaking work that some of the most optimistic (and most pessimistic) specialists expect seems highly improbable.
I think all over that this is hallmark of a very nascent field only just beginning to set its proper foundations, and one where some of the town planners are so excited about debating the value of building a monorail in our small village that they've failed to notice that we haven't invented fire yet and every time there's a strong wind all our stick huts fall down.
Thanks Maxi! The more time I spend working in the field and the more I educate others about ML the worse I realize the definitions issue is. I also realize the scope of impact the problem has. It's very difficult to communicate with others (especially those new to ML) when their definition of a term is entirely different from your own.
Super cool your article
Thanks Pedro!
Was referred this by a friend who knows well my mania for 'The Definiton Problem' in AI.
My experience maps onto yours pretty directly in respect of the definitional confusions I encounter within applied AI - which naturally get worse the further out you go from a centre filled with engineers on more intimate terms with the tech - but I think this is also a much wider issue in theoretical AI too.
Just as there's a lack of terminological definition in ML vs. DL, to use your for-instance, there's a lack of terminological consensus in AI research about what, for instance 'intelligence' (/'learning'/'consciousness' etc.) actually means. In most respects, participants in these sorts of conversations don't even realise that their definitions a. differ and b. are not at all rigorous in and of themselves. Shy of such essentials, the nearness-at-hand of the pathbreaking work that some of the most optimistic (and most pessimistic) specialists expect seems highly improbable.
I think all over that this is hallmark of a very nascent field only just beginning to set its proper foundations, and one where some of the town planners are so excited about debating the value of building a monorail in our small village that they've failed to notice that we haven't invented fire yet and every time there's a strong wind all our stick huts fall down.
Really good piece.
Thanks Maxi! The more time I spend working in the field and the more I educate others about ML the worse I realize the definitions issue is. I also realize the scope of impact the problem has. It's very difficult to communicate with others (especially those new to ML) when their definition of a term is entirely different from your own.