I'm extremely skeptical that this works a fraction as well as it claims. Almost all of the information it provides is astrology-like well-meaning doublespeak that applies to everyone at one point or another. There's even a name for it - http://en.wikipedia.org/wiki/Barnum_effect.
That probably won't stop businesses from wasting truckloads of money on their own automated soothsayer, though.
Edit: Make sure you consult their physical terminal before emailing that big VC! http://imgur.com/UMIlcXz
Yes... To make a generative model that can outperform humans in a social setting? I think it is technically feasible to do basic sentiment analysis, but the generative model... sorry, I think that NLP tech is just not there yet.
On the other hand, I think it is a good idea to have a sentiment analysis system in place that advices users on overly negative sentiment in their comments or e-mails. I wish there was one already integrated into YC comments and into Gmail. Potentially can make life quite a bit nicer actually.
P.S. like your wording 'automated soothsayer'. Funny that the next logical step would be wasting money on 'automated soothslayers' ;)
Sentiment analysis to avoid overt negativity or rudeness is a great idea. It might prevent people like me from commenting too often, and I could see some perennially-depressed Reddit/OkCupid users getting a useful heads up from it as well.
- Give some subjects the advice from this program, and some subjects advice from a knowledgeable co-worker of the person in question (PiQ). (Make sure they don't know which the advice is coming from.)
- Find some way to measure the "success" of the subjects in question -- maybe have them and the PiQ give a subjective rating of the interactions, which should be good enough.
- See if there's a significant difference between the two.
If all such advice is unhelpful, or this service is no better than astrology, you'll see no difference.
An even easier test - give subjects the advice for a different person than the one they're writing to. If there's a similar success rate, then the personality advice is meaningless.
Crystal does have a couple cool things going for it, though - for instance, they're in an excellent position to improve user writing quality by presenting standard stylistic advice (don't use passive voice, don't re-use words, etc) as advanced AI clairvoyance.
> give subjects the advice for a different person than the one they're writing to. If there's a similar success rate, then the personality advice is meaningless
Not quite. This would mean that the personality advice is unrelated to the personality of the target. But it could still be good advice! It's quite possible that a machine that always gave you the same advice, no matter who you were writing to, would still improve your results. (What if the advice was "Are you drunk? Are you sure you want to send this?")
This is similar to a criterion that I once proposed when a personality test was given to the employees at my workplace: Would I recognize myself in the report if I found it on the mailroom floor without my name on it? You could take the reports, remove the names, and let people try to guess which person belonged to which report.
To make an analogy, imagine finding a room full of un-labeled bottles of chemicals, and an inventory of chemical names. Could you match the chemicals to their names?
My immediate thought is applying this to okcupid messages. Okcupid can figure out what kind of messages is likely to generate responses for each person, and can probably charge the sender for the ideal writing style recommendation.
I think that's against their long-term interest. For traffic to keep coming, they need either new users from good word-of-mouth from people who had good experiences, or repeat usage from people who were happy enough with their past experience even if the relationship didn't last forever. Explicitly helping people be manipulative in getting to an in-person meeting between probably-not-actually-natural-matches seems counterproductive.
I worked for a dating website years ago and you might be amazed about their long-term interests..
If you join a dating site and fail, they lose one customer.
If you join a dating site and are successful, they lose two.
The goal became determining good but not great matches and notify the user regularly, especially if they hadn't logged in for a while. It was basically a game of giving false hope to get them to keep their accounts active and the dollars flowing.
When you try it once or twice, it looks quite cool because you see a very detailed description of exactly what the person wants to hear. Then you do it a few more times and you realize that the detailed descriptions are actually a nice piece of copywriting that is re-used over and over again. Some of the magic is lost once you realize that the system is not really as precise as the copywriting implies.
Wired's writing is fine, it's Crystal that's creepy. An unknown proprietary algorithm tells us what to write to our friends? Soon they'll just write for us. Does an advertiser want 1 million people to casually mention something to their friends today? Have the algorithm work it in!
Crystal gives companies a new and more powerful tool for spreading their messages. Of course the first use is always innocent, but if this tech caught on (I can imagine in some form it will) then there is a lot of potential here for PR firms to interject their messages.
EDIT: I should add, I don't see Crystal as bad. I see it as inevitable. People have been blindly sharing private data on the web for a decade because no one used it? People will start using it.
> Wired's writing is fine, it's Crystal that's creepy. An unknown proprietary algorithm tells us what to write to our friends?
To be fair, I think it's more aimed at more casual acquaintances whom you haven't gotten to know as well yet.
> EDIT: I should add, I don't see Crystal as bad. I see it as inevitable. People have been blindly sharing private data on the web for a decade because no one used it? People will start using it.
That's nothing - I've been intentionally, publicly, sharing how to effectively communicate with me for a decade - and information about how I communicate.
I can see the ick factor in giving e.g. your local neighborhood spambot this power - but I want to effectively communicate with people. I want to be effectively communicated to. Hell, I even want to be effectively advertizied to - it means they've picked up on how to avoid annoying me and serve my interests (e.g. I click a lot of webcomic ads.)
I have had so many conversations in so many chatrooms where conflicting or just plain ineffective communication styles lead to frustration on one end, or hurt feelings on another. I extol the virtues of reading various articles about effective communication - "How To Ask Questions The Smart Way" [1] and "What have you tried?" [2]. I've even written a few myself to explain my at times abrasive communication style - extoling it's virtues, defending my choice, and just plain explaining my rationale.
It's a great book where someone basically develops Crystal except it can compose messages too. Then it realizes that humans don't have to be in the loop at all..
>it's Crystal that's creepy. An unknown proprietary algorithm tells us what to write to our friends? Soon they'll just write for us. Does an advertiser want 1 million people to casually mention something to their friends today?
well, another, say Quartz, app would read the message and decide to write back through for example the same Crystal. Complete automating of social life of automatons. While they are at it, we can socialize on our own...
Actually, something akin to that is already being done. Take a look at this http://www.persado.com. The co-founder is a guy from Upstream (a Greek mobile marketing company)
> Wired's writing is fine, it's Crystal that's creepy. An unknown proprietary algorithm tells us what to write to our friends? Soon they'll just write for us.
When you email ian cuss like a sailor and spit a lot. Challenge him to a fight in the first three sentences. Tell him star wars > star trek to throw down the gauntlet. As him about his mother.
The only way this would be a killer feature in my book is if it analyses my entire inbox to generate the result. It would be more realistic since it is actually based on email and not published writing samples, which are not always indicative of communication.
I just signed up and it has the option to suck in your gmail/outlook and provide you a "communication style" profile that's 1) definitely accurate at least for me 2) kind of cool to see a la personality style quiz.
So if you email me or anyone else who has done that it will I assume incorporate that.
Weird. I was just reading William Hertling's Avogadro Corp (http://www.amazon.com/Avogadro-Corp-Singularity-Closer-Appea...) and the premise of that book is that someone builds this product exactly. Unfortunately, it very quickly goes all Skynet/HAL on the world.. oh well.
That's the problem with the increasing speed of technological advance, it makes it impossibile to enjoy SF as by the time you are halfway through the book it's already happened in real life!
Heh, Hemingway is a cute idea, I had not seen that before.
It seems a bit miscalibrated though. I tried pasting in the current front-page article from The Guardian (http://www.theguardian.com/world/2015/apr/24/obama-drone-str...), and it becomes almost completely red ("27 out of 37 sentences are very hard to read"). If it categorizes ordinary newspaper prose as "very hard", I wonder what kind of audience it is aimed at.
It's classic Strunk & White nonsense, the idea that simpler (even to the point of being simplistic) is always, always, always better, founded on the unfounded idea that removing words is the only way to linguistic correctness.
Eh, I think the app is a good first pass for many people. Use it and the Up Goer Five editor as a way to generate ideas for how to shorten and increase readability of your work. I find it very useful when using scientific writing in order to recheck your work for ease of communication.
The augmentation shows you the personality type of the person you're talking to and suggests the best way to get them to do what you want, cajole/reason/appeal to authority, etc.
Not only is the coffee email example an animation, but it's tied to the scroll bar. This make no sense. I was wondering why the text was cut off before I realized what was going on. The images are starting to scroll off the top of the page before it shows me the entire text!
> That of course is the dream implicit in all this: A button that sends the perfect email every time. Indeed, a number of artists have explored the contours of this queasy future in recent months. A browser extension by Joanne McNeil fills emails with exclamation points and smileys, automating the “emotional labor” required in today’s cheerful correspondence. Lauren McCarthy and Kyle McDonald took these ideas a step further with Pplkpr, an app that uses biometric signals sort the real life acquaintances that invigorate you from those that aren’t worth your time.
These are great. Anyone know of any other similar parody apps?
The sceptic in me says that what the average person writes on the web that's easily and accurately traceable to them says relatively little about their preferred email communication style (if the algorithm had access to their "inbox" and "sent items" it might be a different matter.)
I'm intrigued by whether the advice it offers focuses on mimicking their own communication style of whether it gets a bit more clever and human, and advises people (e.g.) to phrase things in a very deferential manner to somebody who gives the appearance of pomposity.
I think this is an incredible idea. IMO, the next generation of high impact startups are going to bring convenience and value to the billions of people using smart but cheap computing devices (phones, for instance). And products like Crystal are exactly in line where it comes to providing a degree of convienience in inter-personal communication. Its all about starting on a problem and continuing to perfecting the solution. If a computer can teach itself to play a game (DeepMind), what's preventing it from learning about other people and their behaviour?
This feels strange but in the movie Avengers, Ultron, an AI algorithm, is born with extraordinary powers that can only be described as Palantir on steroids. Palantir itself is doing amazing work and solving really hard real world problems for Enterprises. Crystal (and others like refresh.io, bright.com, relateIQ) are doing it for the end user. And its pretty evident that something like this has a massive market potential (judging by acquisitions). With the amount of data flowing in the emails, on the social networks, search engines... it must be entirely possible to judge someone with unerring accuracy (the proverbial holy-grail for advertising systems). Mining of information from these signals could help solve a variety of problems-- like helping you find a recruit, a co-founder, a roommate, a life-partner, a friend that shares interests and likes, help create communities and then expand or evolve them, teach you empathy, help you write better, help with thinking... I can see something like this being central to social networking in general. I think Palantir and other companies have demonstrated that these and other similar problems are very much solvable.
"Add your name to the waiting list, and we'll send you an invite when we are accepting more users."
Looked interesting in an abstract kind of way, but can't be arsed to play hard to get. Anyone got documentation on these "64 personality types" and the underlying theory?
It would be interesting to see what happens when everybody in a company starts using this software (mails as only source of information). After some time it would probably lead to very few dominant communication styles. Reminds me of translation party.
I don't know about his boss, but when I read that he responded "Done. Absolutely. It's taken care of." when offered the story, I knew I was reading one juicy story! One he was super-excited to write!
I stopped reading shortly after he let the cat out of the bag that that wasn't his actual response to the stated story.
I believe this algorithm must (partially) base the sentiment decision on the first name of the target. A couple of my friends with an unusual first name did not get a result, even though I know they are on LinkedIn. Also, very common names (Dan, Matt, etc.) come up with the same generic message.
I've been using this now for a few days and it's freakishly good at its knowledge of people. You fear that it's going to be a bit astrological but then you try it on some of the people you know and it's incredibly on point (and the advice varies a lot).
Just tried this out and it seems to be pretty accurate for both me and one friend of mine. Granted that's not a big sample size, but it also has confidence levels.
Anyone having a working account and not just getting wait-listed? Would love to see some examples for comparison, e.g. the same message tailored to Horowitz vs PG?
That probably won't stop businesses from wasting truckloads of money on their own automated soothsayer, though.
Edit: Make sure you consult their physical terminal before emailing that big VC! http://imgur.com/UMIlcXz