I think of bias as an irrational preference. There's an interesting and long list of biases on Wikipedia. We're all biased! It's necessary for efficient decision-making.
Algorithms that make no a priori distinctions between input variables are provably unbiased. That includes pretty much all of computer science. Remember this next time you see a story about a supposedly biased algorithm.
-
-
Tämä twiitti ei ole saatavilla.
-
- Näytä vastaukset
-
-
-
But neural networks can be trained with biased datasets. Maybe that's what they mean?
-
Tämä twiitti ei ole saatavilla.
- Näytä vastaukset
Uusi keskustelu -
-
-
Provable assuming non-hetereoskedasticy. But ML algorithms can be biased against (or towards) groups with higher variance.
-
I often wonder how much of the resistance to ML ethics questions using "AI can be racist" as shorthand for "training data can be skewed, which produces biased outcomes which may conform to racist societal biases, or further enforce those systems of oppression"
- Näytä vastaukset
Uusi keskustelu -
-
-
Does that mean non parametric models and algos too ?
Kiitos. Käytämme tätä aikajanasi parantamiseen. KumoaKumoa
-
-
-
How disengenious. When people talk about bias in algorithms, they're talking about the biases of the trainers, data sets and inputs which we can reasonably expect to have bias.
-
Based on your logic, the algorithm that tagged two black people as gorrilas as bias, is not actually bias. If your thinking stops at, yes that algorithm is NOT bias, ah nevermind
Keskustelun loppu
Uusi keskustelu -
Lataaminen näyttää kestävän hetken.
Twitter saattaa olla ruuhkautunut tai ongelma on muuten hetkellinen. Yritä uudelleen tai käy Twitterin tilasivulla saadaksesi lisätietoja.