Not completely true. Spam is relatively easy to block because you have a complete copy of the communication to work with and not just the envelope.
In the spam world, honeypots are set out. If an email is sent to a honeypot (a fake address), then it is, with high probability, spam. If the email is sent to multiple honeypots, you can be sure it is spam. If a variety of humans mark the email as spam, you can be reasonably sure it is spam. In principal, you can now compare some signature of an email with known spam messages.
These techniques can be used with phone calls, but there is less data to work with.
The approach that applies a turing test to each incoming phone call will filter out robocalls nicely. And it increases the costs of humans calling enough to filter out many of them. For persistent humans, I like the approach of having the sender send a token payment to the receiver which the receiver can easily refund.
In the spam world, honeypots are set out. If an email is sent to a honeypot (a fake address), then it is, with high probability, spam. If the email is sent to multiple honeypots, you can be sure it is spam. If a variety of humans mark the email as spam, you can be reasonably sure it is spam. In principal, you can now compare some signature of an email with known spam messages.
These techniques can be used with phone calls, but there is less data to work with.
The approach that applies a turing test to each incoming phone call will filter out robocalls nicely. And it increases the costs of humans calling enough to filter out many of them. For persistent humans, I like the approach of having the sender send a token payment to the receiver which the receiver can easily refund.