[praseodym]

Just as scary as having Google filter out mail that is totally not important, i.e. spam?

In Gmail you can (un)label email as being important, which will then train an algorithm just like a spam filter. Besides that, it will learn to recognise email you reply to often as being important.

I very much like this feature; I set up the Gmail app on iOS to only send me push notifications for important mail. For me that strikes a good balance between no distractions and not missing out on important email.

[lmm]

Yes, spam filtering is scary. The first thing I've done with every email account I've used is turn any black-box spam filtering down to the minimum (gmail was the first that wouldn't let me turn it off entirely, and it took a long time before I trusted gmail for anything important for just that reason).

Filtering into "email that I want a push notification for" versus "email that I'll read in a few hours when I check" makes sense. What I don't get is people who are using this as a way to divide into emails they'll read and emails they won't read. Given that genuine spam is so rare these days, any email that gets as far as you gmail "unimportant" inbox is an email that you in some sense requested; if your inbox is full of mailing lists and newsletters that you don't read, wouldn't unsubscribing be a better route than filtering them out?

[raverbashing]

Yes, and it's been really due in gmail https://www.joelonsoftware.com/2005/03/29/the-road-to-fogbug...

http://onoffswitch.net/fogbugz-priority-prediction-naive-bay...

[stephenr]

Spam can be removed by objective filters, that simply classify the email against known patterns/rules. Who you are or what you like is irrelevant to them.

This is not that: it's you, giving a giant personal information sponge, a bigger tap.

[loup-vaillant]

> Spam can be removed by objective filters,

Those "objective filters" prevent me from sending email from home (I have to relay through a non-residential IP).

I receive about a dozen spam email per day (with occasional surges and lapses). My server accepts everything, and a simple local filter from my mail user agent (Evolution or Thunderbird, mainly) let few through, and false positives are very rare.

I'm not sure why the giant providers need to work any differently.

[Yetanfou]

A dozen spam messages per day? Lucky you. I did a tally on yesterday's harvest on my server and found the following:

- 57 rejected messages designated as spam by SpamAssassin

- 137 greylisted messages, most of which will end up being spam as those addresses which I communicate with regularly will be in the whitelist.

- 181 connection attempts blocked at the gate due to protocol violations (most of them due to fake HELO, usually trying to connect using my own server's FQDN)

- 144 delivery attempts blocked at RCPT due to the use of blacklisted recipient addresses. This is why using recipient-specific sender addresses makes sense when communicating with commercial, organisational or governmental institutions: it makes it possible both to track down who leaked or sold addresses to spammers as well as to block those addresses entirely.

This domain has been handling mail for close to 23 years now, the server is used daily by about 8 people, it also forwards mail for a few others.

I generally don't see more than one or two spam messages per week in my actual INBOX.

[loup-vaillant]

Ah, 8 people. If we remove the protocol violations, we get (57 + 137 + 144) / 8 = 42 spam message per person per day. Between 3 and 4 times my amount. I may be lucky, but frankly that doesn't sound extraordinary.

I also get no more than 1-2 spam message per week in my inbox.

[halflings]

What new giant personal information are you giving, really? If you use Gmail, you already trust them with your emails and the interactions you have with these mails.

This features makes the experience better for users without having to give any additional information.

[pferde]

Yes, if you're using Gmail, you've already lost, but this IMHO inappropriately affects your workflow by steering your attention to, or away from, certain e-mails. To me, it's potentially no different from what Facebook was caught doing, and in fact, if done carefully, it is an authoritarian's wet dream.

Yes, you can (supposedly) train the "importance" filter, but how many people will do that?

[joshuamorton]

Literally everyone? It's trained based on the emails you respond to.