> Machine learning algorithms are able to accurate identify spam
Nope. Not even close.
> and itʻs not because they are able to tell itʻs about Viagra or mortgage refinancing
Funny, because they can't even tell that.
Which is why mail is being ruined by google and microsoft. Yes you could argue that they have incentives to do just that. But that doesn't change the fact that they can't identify spam.
My experience has been that Google successfully filters spam from my Inbox, consistently.
I get (just looked) 30-40 spam messages a day. I've been on Gmail since the invite-only days, so I'm in a lot of lists I guess..
Very Very rarely do they get through the filter.
I also check it every couple of days to look for false-positives, and maybe once a month or less I find a newsletter or automated promo email in there for something I was actually signed up for, but never anything critical.
You’re talking way too hyperbolically to take seriously.
Yes, GMail does, in fact, “have a clue.” They do pretty well. They’re not perfect, and I have specific complaints, but to pretend they’re totally clueless and inept discredits anything else you’re saying.
Just as saying that machine learning can identify spam discredits anything else ex-reddit CEO says.
I'm sure gmail have a clue from their point of view, but those doesn't align with mine (nor, I'd argue, most of their users). Their view also as a coincidence happens to strengthen their hold on the market but who cares?
Nope. Not even close.
> and itʻs not because they are able to tell itʻs about Viagra or mortgage refinancing
Funny, because they can't even tell that.
Which is why mail is being ruined by google and microsoft. Yes you could argue that they have incentives to do just that. But that doesn't change the fact that they can't identify spam.