Lessig Wagers His Job On Anti-Spam Theory

kien writes "Lawrence Lessig is betting his position at Stanford on his anti-spam legislative recommendations. From his blog:'First the analysis: Philip Jacob has a great piece about spam and RBLs. The essay not only identifies the many problems with RBLs, but it nicely maps a mix of strategies that could be considered in their place. But, alas, missing from the list is one I've pushed: A law requiring simple labeling, and a bounty for anyone who tracks down spammers violating the law. Here goes: So (a) if a law like the one I propose is passed on a national level, and (b) it does not substantially reduce the level of spam, then (c) I will resign my job. I get to decide whether (a) is true; Declan can decide whether (b) is true. If (a) and (b) are both true, then I'll do (c) at the end of the following academic year.' The Declan referred to in point (b) is Declan McCullagh." Update: 01/07 02:45 GMT by T : Speaking of whom, here is Declan's acceptance of Larry's bet.

something missed

[neildogg]

They missed the link to his idea [cioinsight.com]

Re:A is A

[odin53]

Lessig is a conservative himself. He clerked for Richard Posner, a famous strongly conservative judge for the 5th Circuit, and Justice Scalia on the SCOTUS.

Re:A is A

[kien]

Here's one Slashdotter who would be happy to see this fraud fade slowly into the good night.

Well, you're entitled to that opinion. But have you read Professor Lessig's book The Future of Ideas [amazon.com] where he academically supports his argument.

It's easy to have an opinion and forecast the future (as you have done). It's much more difficult to have an educated opinion and forecast the future with references (as Lessig has done).

Here's one slashdotter that smells a troll. :)

--K.

Re:NATIONAL law will stop third-world spammers?

[kien]

The problem is it's being addressed on a national level.

That's a valid point. It leads to what I personally call a "slashdot paradox". I'm outraged that a Russian programmer (and then, the company that employed him) was prosecuted here in the US for software that is legal in Russia. Yet if Prof. Lessig's law is passed in the US...paradox. It could be argued that laws passed in the US have typically been adopted in one form or another around the world (which sucks...ref: DMCA) but that could be countered by the jurisdictional nightmare that the RIAA/MPAA have run into while trying to prosecute Kazaa.

The blessing and (for right now at least) the curse of the Internet is that it globalizes the public commons. We're only now beginning to confront all of the issues that are raised by this fact.

--K.

Re:Please resign now

[kien]

He simply refuses to understand that we are quickly entering into an age where either all information will be controlled or all information will be free. Information is so easy to copy, modify, and manipulate - there can be no middle ground.

Actually, after reading his book [amazon.com] I'm more inclined to think that he understands the issues at stake on a different level.

--K.

Re:First problem with this solution:

[sfe_software]

Just as an aside it's "Bayesian".

Crap, I've made quite a few posts with the double-S up to now... ugh! Gotta love Google Suggestions though :)

...the success of Bayesian Filtering is because it is rare:...

I'm not sure about this. The success of Bayesian filtering (rather, Classification) is that it learns what you, the end-user, considers SPAM. How does a SPAMmer learn what you, personally, don't consider to be SPAM?

Currently I get a TON of SPAM with "jm4n" in the subject (my most common email username), and often it sounds like a reply to a personal email ("thanks for your email! Look at my web site and see me naked!").

But the point is, since it learns, and it's completely based on individuals (rather than some generic description of what constitutes SPAM), it works rather well -- better than you might imagine. I suspect this will work in the future.

I also mentioned:

If some implementation were to add common word-groups instead of just word occurrances, it might even be more rock-solid

Currently Bayesian classification is extremely simplistic. It classifies word-counts, and figures if the word "Viagra" never appears in "real" mail, and often appears in "Junk" mail, then this is a key indicator that this is SPAM.

If this were extended to word groups (say, groups of words that appear together frequently, like "penis enlargement" or "work from home"), it could even be more effective. I'm tempted to work on some Perl scripts to implement this even further than what Mozilla does (and, being on the server-side, stop wasting bandwidth on my DSL; not to mention multiple-client compatibility)...

Re:First problem with this solution:

[ergo98]

I had a long winded reply regarding false positives and what they represent to even the best filtration (i.e. what happens when your filter is attuned to emails between you and your buddies, and suddenly a proposal comes in from an employer, or a partner, or a customer? This single lost email could be incredibly damaging) when I noticed this [jerf.org] page that says it eloquently and thoroughly.

Re:Sting the bastards into oblivion

[Anonymous Coward]

As someone who hypothetically works at a Tier-1 provider who hypothetically has many spammers buying bandwidth, I can tell you that any spammers relaying off a 1.5 Mbps pipe are small fish. Most big fish spammers purchase 10-100 Mbps of pipe to the core at a time and use all of it. They don't bum bandwidth off other people's relays.

Of course, this highlights another point I think most people are not aware of. For network providers, spammers can be big business. Many of the bigger spammers purchase fatter pipes than many of the big technology-centric companies out there. As much as ISPs despise the headaches spammers bring them, spammers are also among their biggest clients. Spammers consume backbone bandwidth at rates that few other businesses do and this is translated into profits at a time when the ISP business is pretty rough. Since most ISPs require spammers to pay upfront for their bandwidth, this is a welcome addition to the bottom line. Obviously there are people sending spammers money or they wouldn't be able to afford these network pipes.

Re:First problem with this solution:

[WatertonMan]

I pointed out Apple's Mail as an excellent example earlier as well. One point I should make. Unlike Bayesian filters that go through your POP server before you do, the fact Apple's is integrated into the Mail client makes all the difference. You can go into the spam folder and if there is a message that isn't spam, you tell the program and it updates the Bayesian tables. If you find one that Mail missed, you just right click and tell it that it is junk mail.

Apple has it set up in a very friendly way. It prompts you initially for mail it thinks is spam. At that point it has preliminary training from Apple plus whatever you give it. At a certain point it figures it is accurate enough and goes off on its own automatically.

As I mentioned, after a little training myself, it has yet to make a mistake. It is amazing what that does to your workflow!

Don't get me wrong. OSX Mail isn't a perfect mail client. It isn't aware of the blockquote HTML tag, for instance. And I hate how it uses a drawer that isn't resizable for its folders. Unlike the old Claris Mail it doesn't have scriptable triggers for certain mail events either. (Of course neither do most mail clients, especially on other platforms) But other than that it really is very, very impressive.

Re:Rubbish

[viscous]

The analogy you are trying to make doesn't work. Refusing to accept your electronic mail is not the same thing as killing you. Killing you is immoral and illegal. Configuring my mailer to refuse electronic mail from your ISP is neither.

I suppose the problem here is the use of the phrase "collateral damage". It is an unfortunate military metaphor which does tend to suggest the analogy you are trying to draw. But it won't work.