[SoftTalker]

I don't understand why spam detection is so complicated. I can tell with high accuracy if an email is spam just by the subject line. I'd think even basic ML could do this very reliably you don't need a bleeding-edge LLM to do this.

Phishing is tricker because it can be very deceptive especially if you're being targeted specifically. But also usually pretty obvious.

[PaulDavisThe1st]

ORLY?

* Are you available? * Paul, can we have a zoom meeting with you on Monday? * Assistance for donation * Greetings!!! * some ideas for you * Refund request * Somethings not working * Manuel Montoya for roof work contractor * proposals for print * Invite Connection

Half of the above are actual spam, half are not. Tell me which is which ...

[maxbond]

I agree with you but for the love of the game, these are the ones I'm guessing are spam:

* Paul, can we have a zoom meeting with you on Monday?

* Assistance for donation

* Greetings!!!

* Invite Connection

* Refund request

[PaulDavisThe1st]

3/5 :)

[maxbond]

Better than Google did today!

[SoftTalker]

If there's any question look at the sender. I delete half my email unread based on subject and sender alone.

[DANmode]

You can tell from your subject lines.

You cannot 100% tell from others’ subject lines,

if you don’t know them personally.

[SoftTalker]

Yes but I also don't have the ability to know what millions of accounts are receiving, unlike Google.

[DANmode]

Neither do they - until hours after you’ve potentially received a spam message in your inbox.

It’s past patterns + live human weighting.

[chistev]

They can tweak the subject to something not obviously spammy.

[plagiarist]

Obviousness in spam is a feature, they don't have to waste effort on people who know better.

[Nextgrid]

This only applies to spam which requires significant follow-up effort from the spammer to respond to potential victims; effectively just 419 "advance-fee" fraud scams.

For spam which only does not require manual effort on the other side, there is no reason to filter out potential victims and all the more reason to make it look as legit as possible to maximize conversion rates.

[takanot]

> For spam which only does not require manual effort on the other side, there is no reason to filter out potential victims and all the more reason to make it look as legit as possible to maximize conversion rates.

Unless there's a trade-off. Saying "respond now or your account will be erased!" doesn't sound very legit. But the number of additional victims the phisher gets by doing probably outweighs the number of more sophisticated victims he loses.