Top Tweets for #Notaduck

Apparently lots of people are stuck in a sort of “Duck Metaphysics”. Meaning: if it quacks like a duck etc => it is a duck. If we cant tell A from B then we can just assume A=B. Even if it isnt, it makes no difference. Which appears to be pragmatic and logical— but it’s actually nonsense and leads to chaos in practice. For example it would imply an early stage cancer we cant diagnose isnt a disease at all, or its irrelevant if its a disease — no need to look for better diagnosis tools, if the person looks healthy then she is healthy. In fact Duck Metaphysics only make any sense in theoretical scenarios that dont exist in real life. Sure, if *any conceivable observer, in any circumstances, with any amount of time* can’t tell A from B, then yes we can just assume A=B. Of course its an untestable property. So saying things like “if it passes the tests we have then it doesn’t matter what you think about the code, it’s correct” — is the **opposite of pragmatism** And for the most famous example: the Turing Test. If a machine can fool people about being a human, then it must be intelligent/conscious. Which, frankly, is a completely moronic idea. Yes Turing was a genius. Geniuses have moronic ideas too. You put a team of 100 people talking to a machine for hours. All are convinced they are talking to a human. That’s absolutely no guarantee it will act in a human way if you put it into a different sort of conversation or simply keep talking longer. It’s the most extreme (and most nonsensical) case of Duck Metaphysics. We know A!=B. Yet we propose they share a very important property (intelligence) we can’t even properly define. We then go on to create an extremely limited test scenario, and if people cant distinguish A from B in that scenario then the property is valid for both. There is no logical connection between the two facts — fooling humans and being intelligent. The implication is assumed. “If it’s very hard to make a machine that fools humans without implementing something functionally equivalent to intelligence — then a machine that fools humans is probably intelligent” Except its presented without the initial conjecture as 100% guesswork. You can test millions of random inputs on a calculator with faulty carry propagation without detecting a single error. Then trust it in a real world scenario and have a catastrophic failure that causes a bank balance to double. Proper testing — white box, skeptical, adversarial — could likely catch it. Naive duck metaphysics wont. If it seems to you something quacks like a duck and walks like a duck— maybe you just have bad eyesight and its a movie projection of a duck. Neither science, nor engineering is about producing convincing fakes. Duck Metaphysics is just stupidity and fraud posing as pragmatism.