▲ 187 points
back
100 comments
If you need to send out 200 personalized letters to sales leads but haven’t got the time to handwrite them yourself — or if your handwriting is, like mine, grotesque — then Maillift will generate them for you, using teams of genuinely carbon-based people.
That's horrible! I don't mind getting a printed letter but if I got a handwritten letter that I later found out was faked I would never, ever do business with that person again. Doing this means the very first step in an attempt to form a business relationship is an act of deception.
This is of course a classic submarine story: http://www.paulgraham.com/submarine.html
Every time I feel like I have come to terms with marketing people as economic actors they manage to plumb some new nadir.
Would you be offended if you found out that a hand-written note you received was actually written by someone's secretary? Because that's pretty standard stuff most people wouldn't consider an ethical breach.
Traditionally if you wrote a letter on someone's behalf (eg they instruct you by phone on what to say and to get it in the mail as fast as possible), you sign is 'PP J. Smith'. PP is short for 'per procurationem' meaning 'by the agency of'. You don't need to know Latin, just that 'PP [name]' means it was sent on someone's behalf.
If I got a letter with no indicator and it turned out to be someone's secretary I wouldn't be offended exactly but it would make me a little wary until I was more familiar with the person and their working habits, eg maybe they have total trust in their secretary, but the secretary isn't familiar with this convention.
Obviously I'm a bit old-fashioned about these things, bt these conventions came into being for good reason.
DirecTV and credit card companies do this to me all the time. I'll receive a letter that looks like some kind of greeting card or invite with no identifier but an address. The font used on the outside mimicks human handwriting, so it looks authentic if you're not really paying attention.
I kinda get excited I'm getting invited to something, I open it up, and its a fucking ad.
I don't understand how people do business with businesses that mislead you like this. I mean, if they think of you so little before you're a customer, imagine how badly they will treat you when you are one.
What about when you get an email from a website you just signed up for that opens with "Hi, Anigbrowl! I saw that you signed up for my site..." and is "signed" by the CEO. If you found out that was an automated email and that the CEO didn't really type it up personally, would you be equally upset?
> Every time I feel like I have come to terms with marketing people as economic actors they manage to plumb some new nadir.
Well, yes, but they do what works. They wouldn't do it if it didn't work. Which means they manage to connect with something deep inside our minds. So, when you make statements about what marketers do, you're really commenting on how the human mind works - a jumbled, impromptu, cobbled-up-in-a-hurry mess (that somehow still manages to put rockets on Mars, by some incomprehensible miracle).
What if instead of employing an outside company these were in house employees?
"Marketing people" is one of the most egregious generalization errors on HN.
It's certainly in vogue to hate on marketers but it is not an identity, marketer is a hat that people wear. Just stop.
Well, you can't.
You can buy a wacom tablet with tilt sensors and make the robot replicate the same stroke. You could add randomness to it too.
To apply different amount of force with the robot you use a spring and you control the force with the distance to the paper, it is very simple to do.
We did exactly that with ABB robots in the University a long time ago.
We used wacom drivers for linux and the wacoms were over 200 dollars or so.
A robot that uses only 3 axis is not "five figures", it is basically a plotter and thanks to reprap you could make your own for over 300 dollars.
Those are probably the ones used by this people, but you can use a robot with more degree axis and you won't be able to differentiate it.
Yeah, fully replicating human hand writing with varying force and slight randomization of letters sounds fairly easy to do. Sounds like a project an intern can handle.
Do you have any links to papers or code that could help get me started on such a project? This sounds like it'd be incredibly fun to build.
3 axis would be x, y and tilt?
Judging from the Robot letter in the article, the easiest way to spot a robot-penned letter would be that specific words are all the same. Look at the word 'the' in the picture, or even the 'th' in 'this'. They're all identical.
The robot seems to be working from just a small pool of characters.
Some sort of randomisation algorithm would work well here, varying the size, weight and slant or jitter of each letter would at least make it appear more random, and harder to spot.
Having said that, if you are someone who works on this sort of invention, please take a long, hard look at what you're doing with your life and consider just how this invention is contributing to the advancement of mankind.
> Having said that, if you are someone who works on this sort of invention, please take a long, hard look at what you're doing with your life and consider just how this invention is contributing to the advancement of mankind.
This is a silly point that gets repeated far too often. You can make that statement to almost every developer and designer here.
Some people are quite satisfied advancing their own lives and that of those who are close to them.
the amount of mechanical engineering, signal processing etc, involved, they could be working on a bigger, valuable in-need problem? even more artistic would be cool...
This is so pointless. People have positive associations with handwritten letters because of the content that goes with it. If marketeers start using handwritten letters they will just ruin that association.
But in the few years before the association is ruined, those marketers that successfully fool people into reading their junk mail will make money.
So? Marketeers didn't pay for that association and aren't really invested in not exhausting it.
Time to invest in wax and a signet ring
As a kid, when I used to write on white paper, my lines used to slope and I used to throw them in the trash and try again. Over time, underneath the white sheet of paper, I started placing another paper that has lines and a margin, so I can see through the white paper and know where exactly I should write. The result was perfectly straight lines and a perfect margin.
I'm not a robot.
> I'm not a robot.
How do you know?
Computers can imitate human handwriting styles as well:
http://www.cs.toronto.edu/~graves/handwriting.cgi
Although a simple way would be to just have a human write the original and have a robot copy it.
Shouldn't marketers try the exact opposite approach?
In this paper about "Nigerian" mail fraud http://research.microsoft.com/pubs/167719/whyfromnigeria.pdf
researchers argue that it's very efficient for attackers to appear fraudulent, because this helps select the most gullible victims (true positives) while weeding out "false positives": people who respond at first, but are not scammed in the end.
From the abstract:
> "Far-fetched tales of West African riches strike most as comical. Our analysis suggests that is an advantage to the attacker, not a disadvantage. Since his attack has a low density of victims the Nigerian scammer has an over-riding need to reduce false positives. By sending an email that repels all but the most gullible the scammer gets the most promising marks to self-select, and tilts the true to false positive ratio in his favor."
Shouldn't marketers do the same? Isn't it more efficient to eliminate non-targets as early as possible, instead of dragging into the process people who will ultimately relent. If so, it would seem preferable to send junk mail that is very explicitly junky.
Maybe it's because they have different goals?
The Nigerian-mail-fraud-scammer needs to get the gullible person to respond. If that's 1 out of 100,000 people, they need to do what they can just to get that 1 to respond. If 100 out of 100,000 respond, they have to sift through the "noise" of not-so-gullible people to find the true gullible one.
In this case, they just want to get you to open the letter and read the sales pitch (the goal isn't to find gullible people).
That site was posted on here before, to which someone quite reasonably offered the alternate hypothesis "what if the scammers sending those particular emails are on average really stupid, as attested by anyone who has ever attempted to interact with them"
If you're sending out direct mail marketing and yet have an ordering process so cumbersome (or service of such dubious legality) that regular folks contemplating buying your product are a problem, your company isn't going to be around for very long.
Very explicitly junky mail works because the target demographic actually do get excited by the additional free services worth $100!!!! and nobody else would care to respond to the mail anyway
It's unlikely that mortgage companies are deluged with bad leads because people thought the handwriting was legit. Weeding out false positives is only an issue if you have a false positive problem.
It was an interesting advertisement for Mailift. I find it interesting to see the lengths that junk mail goes to in order to get read. At least at my house opened or not the conversion rate is zero. But clearly that isn't the case elsewhere.
Other than the company is getting better at mimicing humans though I didn't get a whole lot out of it.
When a volunteer signs up at Social Coder I send them a welcome message.
Here's my confession: I will probably have sent that exact same message to someone else in the past.
At the moment I personally click on the "new message" button, fill out the name, paste in the message, double-check I'm sending it from a socialcoder email address, and then I send it.
This "welcome" process would be trivial to automate and yet I hesitate to do so.
Robots (and scripts) also have a way of blundering into situations that humans would avoid but this isn't why I'm reluctant to automate.
When it comes down to it I don't want to take the human (me) out of the loop. It would feel like I'm duping the recipient who might naturally expect a human connection rather than a recorded message.
If the volume was very much higher I might - apologetically - automate the welcome message. It might not make any difference to the volunteers signing up, but it would make a difference to me.
"The Dots on The 'i's" and "The Rounded Right Margin" can both be fixed by having a human generate the strokes for the original form letter, like an Autopen with RAM. Write once, then make 30,000 copies of your exact handwriting with your handwriting robot.
Reminds me of the small company the protagonist works for in "Her".
To expand on this, the protagonist writes personal letters for other people, not because of penmanship but because of wordsmanship.
I recently had a horrible experience getting a home mortgage, and when invited to leave feedback I did so - politely, but sparing no feelings.
In response, I received a form-letter generic apology with a stamped signature. The numeric code at the very bottom of the letter (and the fact that it didn't address any specific issue I'd raised) gave it away, though I had to squint to determine that the signature wasn't legit.
I don't know which bothers me more: the fact that they didn't actually care enough to write a genuine response, or that the get enough disgruntled customers that they have a standard apology letter ready to go.
Fascinating - I had no idea such a machine existed. Here is a video of one in action - https://www.youtube.com/watch?v=k49ncf60zN8#t=16
It's not exactly the same, but pen-wielding robots existed for a long time(earliest examples are almost 200 years old!) - US presidents have been using them for a long time to sign official documents:
One of the things that readily annoys me is the "Important Information About your Current Account" letters from a credit card company that are just purchase checks inside. I do tend to open them, just in case; but have never actually used one. I don't think there is necessarily any trickery going on here, but it certainly trains people to ignore mail from them.
The mortgage companies are worse, in that they plaster the name of your current (or former) lenders on them; even though they have no actual association. These are usually easy to differentiate from real mail though.
Similar machines have been in use for a long time to sign letters https://en.wikipedia.org/wiki/Autopen : US presidents (and LDS presidents) have long been fond of them. Presumably extending their use to whole handwritten letters is newer.
Then there's something even older: https://www.youtube.com/watch?v=laJX0txJc6M
I think neural networks (RNNs to be specific) have the potential to do this - coupled with these handwriting bots it would be pretty hard to tell which is which. See the research of Alex Graves http://www.cs.toronto.edu/~graves/handwriting.html
This type of research is also pretty amazing for speech synthesis.
This is news? The technology is 249 years old ... :)
Realtors have been using these lately. Had to let my old house go back to the bank in a deed-in-lieu earlier this year, but within days of the MLS listing expiring with our realtor, we started getting mail forwarded to our new house that looked handwritten.
All of them were offers from realtors wanting to re-list our house and try to sell it, and it wasn't until I got two identical letters, in identical handwriting, from two separate realtors, that I realized they were computer-generated.
What about stamps though? Real letters from real people have stamps that are glued on in an odd position; junk mail (or any automated mail) has the postage directly printed on the envelope.
It's certainly possible to automate the sticking of stamps, but won't all those operations needed to "look real" increase the cost of junk mail in unsustainable proportions? Which is the point of course, from the receiver/victim point of view...
This machine seems a standard automation for pre-email era.
Most of the e-mails sent by me now and in the past years at my work in HR, have been automated by me while retaining a good amount of "personalization" for most of the non-priority e-mail replies which helped me highly for saving time for my start-ups. You will be surprised at the amount of "personalized looking" e-mails received that are actually automated.
The best way to tell is to check if the postage says "Pre-sorted Mail" or if they actually spent money on first-class postage to send it. ;-)
If the amount of energy and money spent on this level of sophistication can be devoted to actual customer service, imagine their retention. After all, it's the retention that keeps your business going. If your customers will spread the word about you too, that's far more convincing than "convincing looking" handwriting.
Well, at least that's one marketing battle I've already won.
I don't read ground mail anymore.
Now, that is progress. Manipulating a corporate droid into closing a deal by sending them a hand-written / hand-botted / whatever letter is serious kung fu.
I think I'll just rip open the letter rather than analyzing the writing. Too many details to look at.
All my letters are either from a robot or my Nan.
Left handed with horrible penmanship skills -- this is wonderful!
Is there a Makerbot equivalent for these machines?