Any example of such things that are too complex for us to understand? why they hard for us to understand? in essence do we learn the same way machines learn, and hence our process of learning aligns (at every abstraction) to that of machines? what is exacly meant by explainable?
The biggest advantage of ML algorithms is that they can learn things that are too complex for us to understand. Demanding that they be explainable deprives us of the vast majority of the good they could do.
-
-
-
He's likely speaking about causal inference, which indeed, is hard to get from ML (particularly deep learning) models.
Keskustelun loppu
Uusi keskustelu -
-
-
ML algorithms should have to be explainable only after the meat-bags that demand they be explainable can fully explain how their own meat-bag functions.
-
Ew, but also: Why should I have to be able to fully explain how I work as a precondition for wanting to understand how something I built works?
- Näytä vastaukset
Uusi keskustelu -
-
-
Like what? Understand tech industry dumbed down consumers and now promotes the illusion of artificial intelligence which is actually in simple terms pattern matching tied to action to take on a match. Can you think of any other way to promote black box devs don't understand?
Kiitos. Käytämme tätä aikajanasi parantamiseen. KumoaKumoa
-
-
-
I think heavy scrutiny of algos before entering a setting like healthcare is important. But I too am worried that human intuition is favoured in healthcare with higher morbidity+mortality due to medicine's inability to relinquish control even with evidence for ML superiority
-
You can argue that it is hard for ML to "relinquish control" in the sense that it is stubborn and unable to see beyond the narrow purview of its training data.
- Näytä vastaukset
Uusi keskustelu -
-
-
Not sure you can validate a system enough to trust it to do anything significant if you have inadequate explainability/interpretability.
Kiitos. Käytämme tätä aikajanasi parantamiseen. KumoaKumoa
-
-
-
That’s a philosophical argument. It’s not that they learn something we don’t understand. It’s just that they find an algo that we can’t come up with ourselves. We are just lacking the tools to make sense of that now.
- Näytä vastaukset
Uusi keskustelu -
Lataaminen näyttää kestävän hetken.
Twitter saattaa olla ruuhkautunut tai ongelma on muuten hetkellinen. Yritä uudelleen tai käy Twitterin tilasivulla saadaksesi lisätietoja.