3 ms·
It is worse then a blacklist. Blacklist is easy to understand, as long as we trust Google (lots of us don't) everything would be fine. With ML, not even Googl
by j16sdiz 8y ago
It is worse then a blacklist.
Blacklist is easy to understand, as long as we trust Google (lots of us don't) everything would be fine.
With ML, not even Google have a full picture of what's going on.
- kevingadd 8y agoAt this point ML to govern things like autoplay or address bar is just whitewashing for biased data. You feed biased training data into an ML algorithm and now it's unbiased! Select your training data set and other parameters such that you get a result you want and you're good to go. The history of machine learning for tech is a history of obvious biases creeping into training data - whether it's a medical algorithm "learning" that background markings are an indicator of disease instead of classifying the tissue it's meant to, or recidivism risk algorithms going off race & gender to the point that they produce actively bad data, or face detection algorithms thinking that asian people are squinting. I don't even think that youtube necessarily should get an autoplay prompt on first use, but it's pretty convenient that ML-based approaches like this are used instead of much simpler approaches. Lots of research is going into creating adversarial data given known ML algorithms, as well. If this address bar ML is running on the client (it'd have to, right?) then it's not hard to do a training run against it to come up with custom tailored URLs/sites to get the ML to classify your attack as good.
- muthdra 8y agoWhile I agree that it is whitewashing biased data, maybe they get good results with new and unseen URLs that try to look like some relevant page using the same tricks as the scam URLs in the corpus.
- zestyping 8y agoReally good point. ML equals diffusion of responsibility.
- pas 8y agoNo, it doesn't. Google is still responsible. The team at G maintaining it is still responsible. An ML solution is a completeness vs correctness trade off. ML can make the blacklist virtually infinite long, whereas a human team would likely burn out (and make more/different mistakes).
- dhimes 8y agoAnd what do you think will happen when they mess up pas.com/signup? Do you think they'll fix that for you? I hope the team at firefox is licking their chops to implement a better and distinguishing alternative.
- laurent123456 8y agoUntil the day a Google domain is accidentally blacklisted. Then suddenly there will be an internal whitelist where they get to decide what goes in.