5 ms·
Well designed ML techniques are always correct 99% of the time. Its the 1% that is the problem.
by rusticpenn 6y ago
Well designed ML techniques are always correct 99% of the time. Its the 1% that is the problem.
- ukj 6y ago99% of the time it works all of the time. Like humans. How do we deal with problematic humans? Retraining or replacement.
- adrianN 6y agoNo, we deal with 99% accuracy in humans by designing all our human-based algorithms to be resilient against mistake (and still mistakes in the end result happen very frequently). This is completely different from the way we have designed our computer systems, because it's way easier to do this with flexible agents like humans than with inflexible agents like computers.
- ukj 6y agoSo, the mistakes algorithms make are different to the mistakes humans make? That’s... equivocation. Both humans and ML algorithms are flexible. That is the point of “learning”. Adaptation.
- Random_ernest 6y agoBut the problems where we apply human labour are vastly different from the ones where we apply machine labour. In (most) tasks where we apply human labour a few errors are tolerated.
- ukj 6y agoThis seems irrelevant. Neither humans nor ML make zero errors. Ceteris paribus, if an ML algorithm makes fewer errors at a task which with low error tolerance - you would use the algorithm instead of the human, no?
- mtrower 6y agoI would expect that might depend on what sort of errors each make, no?
- bryanrasmussen 6y agowhat's the promote option in this analogy?
- ukj 6y agoML took your job.
- fiblye 6y agoWith humans, we can ask what went wrong and fix the problem. With black box algorithms, we throw in some new training data and just hope it's enough.
- andybak 6y ago> With humans, we can ask what went wrong and just hope we've fixed the problem. > With black box algorithms, we throw in some new training data and just hope it's enough. A small wording change and they're equivalent again. To some extent human beings are also a black box - with some very peculiar failure conditions and side-channel weaknesses.
- bufferoverflow 6y agoNo, we don't do anything against the humans who are 99% correct. 99% is good enough in most real world cases.
- rusticpenn 6y agoThat is not practical when you expect the system to be correct 100% of the time and make decisions based on that. There are many situations where this is critical. You would never want your car's safety system to be correct only 99% of the time.
- AlphENsign_Tech 6y agofrom the volumetric standpoint, the 99% is their huge specially curated training and test data.