Locations
“I’ve seen things you people wouldn’t believe. Attack ships on fire off the shoulder of Orion. I watched C-beams glitter in the dark near the Tannhauser Gate. All those moments will be lost in time, like tears in rain. Time to die.”
Stirring, poetic prose. Especially from a robot.
Science fiction gives us many examples, like this one from Blade Runner, of computers and artificial life forms expressing themselves in terms that humans would be proud of.
What about science fact? Well, we all know that Zooey Deschanel, Samuel L. Jackson, and John Malkovich are able to hold fruitful (or, in the last case, creepy) conversations with Siri, but they are a little one-sided, and there are doubts about how accurate the depictions are.
Narrative Science takes things a step further. As this recent article in Wired recounts, Narrative Science trains computers to write stories—and not just any stories, Kristian Hammond, one of the company’s founders, believes his computers are capable of writing a Pulitzer prize-winning story within five years.
I don’t want to re-tread too much of the Wired article, as you can read the article itself for details (and I recommend that you do!), but one section stood out to me:
As the computers get more accomplished and have access to more and more data, their limitations as storytellers will fall away. It might take a while, but eventually even a story like this one could be produced without, well, me. “Humans are unbelievably rich and complex, but they are machines,” Hammond says. “In 20 years, there will be no area in which Narrative Science doesn’t write stories.”
Now, I don’t know much at all about algorithms, but I do know how to recognize a metaphysical belief presented as fact when I see one—in this case, that human beings are “machines,” no more than the sum of their parts.
If that is the case, then yes, there is nothing that a human can do that can’t be replicated; with enough processing power, the correct parameters, and the right math to connect those parameters, a computer—or an android—can do anything we can do, and better. Our friend Roy from Blade Runner would be a Replican, not a Replican’t.
But the idea that human beings are no more than the sum of their parts is not a given. Philosophers have been wrestling for millennia over the question of whether humans are just the sum or more than the sum of their parts. The question is of supervenience—do the lower-level properties of a system determine its higher-level properties?
In this context, the question is “Does the storytelling capacity of human beings supervene on our neural pathways, our knowledge of the rules of composition, and the amount of data we have access to?”
Narrative Science is not alone in thinking that it does. Epagogix uses an algorithm to predict box office takings of proposed movies, crunching numbers on elements including the script and plot. If even creative pieces like film scripts can be coded into their component parts, then case study writers in particular had better watch out—as they are tasked more and more with dealing in data (hard metrics as opposed to soft benefits to the individual).
However, I would argue that there is more at work in a great script, case study, or any piece of writing that conveys a story—that they are more than the sum of their parts. There is something irreducible about them. When it comes to case studies, the key is that they are records of testimonials—one human being telling another human being about how a product, service, event, or solution impacted his or her life. Metrics, and other definable parts like quotes and proof points, will undoubtedly feature strongly in the story, and the text should adhere to the rules of composition and to brand and legal requirements. But a great case study will also have more than that: an intangible—the feel of a true human connection.
Computers might be trained to mimic that, but if we are more than the sum of our parts, they can’t actually experience it, and that will ultimately be apparent.
Or will it…? Was this blog post actually written by a computer?!
Let me know what you think, either in the comments or via twitter.
Fantastic post David! I think you bring up some interesting points. You’re obviously on the side of the humans and I’m going to take the side of the machines. (Maybe I’m a T1000.)
I agree that machines will never be able to replicate the entire experience of what it means to be human, but I think that they will get 95% of the way there and for more applications, that will be good enough. What I think will happen is that androids/robots/machines will do 95% of the “hard thinking” and humans will become the polishers. We will review and supervise the work of the robots. When an algorithm writes a story and describes two people caring for each other, humans will polish it to be called love.
I know machines are scary and that we don’t like the idea that something is better than us, but we better get comfortable with that idea. See the robot that never loses in rock-paper-scissors. http://www.slate.com/blogs/future_tense/2012/06/27/rock_paper_scissors_robot_from_ishikawa_oku_laboratory_wins_100_percent_of_the_time_video_.html
Why is it bad that machines will be able to write, sing, dance, cook, weld, fabricate, analyze, and do a million other things better than humans?
Humans won’t become obsolete, but our roles will change.
Good comment Brian!
Believe me, I’m no Luddite , so I am very happy when machines enhance what humans can do (e.g. can someone please create/ point me to a program that will maintain my twitter account as if I am creating the tweets!
).
I guess I just prefer to think of them as enhancing what humans do, rather than replacing them (even if, in practice, they do replace a large number of human workers) – as replacing doesn’t honor the unique human element that humans bring to activities they engage in.
Creating lots of automated content to cover things a human writer can’t or won’t – enhancing.
Winning a Pulitzer – replacing
Hi David,
Thanks for taking on this timely and fascinating topic. I was at a social gathering recently and chatting with a friend from The Netherlands who works as a developer for a large technology company. I told him that it didn’t make sense to me that developers make more than writers because you can learn to code C++ language at, say, age 20 — and do it with elegance. But to master a human language — and English is a difficult and nuanced language with a vast vocabulary — you have to start as a child. (Having said so, author Milan Kundera left his native Czechoslovakia in 1975 to live in exile in Paris and he began writing novels in French, so it is possible — though French is much easier to learn than English. Plus Kundera is a human being, and the most famous and well read of his novels were written in Czech.)
Maybe someday computers will do most of the writing, but frankly it’s hard to imagine. If it comes to that, I’ll learn C++.
Cheers,
Molly Dee
It’s a fascinating topic, with both technical and philosophical angles. A very well-informed machine might have been able to write this article – but could it have made the Blade Runner connection, and pulled in the opening quote? We are very complex connection processors, we humans.
Thanks Norm! A machine might have tried to but falied – but that would be OK (knowing reference to previous blog of yours)
I agree we are complex, but I do think a machine could come up with something similar. Would it be exact? No, but people aren’t exact either. With the amount of data out there and the number of times Blade Runner is referenced in relation to robots, machines, and science fiction, it isn’t impossible to think that a machine could make the connection.
Bravo to David. This is a smart and provocative post. As a professional human writer—who likes to think he is a bit more than the sum of his parts—I am gladdened by David’s defense of the human connection, which I like to think is the very heart of customer engagement, a heart I’m not sure we can count on algorithms to provide. Along those lines, I think there are a couple of things worth pointing out here.
First it is important to note that the stirring quote David begins with was not actually written by a robot; it was written by a human, a particularly sensitive one. I’ve been working at it seriously for 30 years and I’m still not sure I could write at the level of Phillip K Dick, so I’m a bit appalled at the notion that a machine could. Whatever. Call it sour grapes if you want, but when Kristian Hammond claims that there won’t be anything Narrative Science can’t write, I’m inclined to say prove it.
David nails the pertinent question: computers might be taught to write prose, but can human experience be automated? If it can’t—and I don’t think so—I believe that failure will be apparent on the page (or screen).
Consider a phenomenon called the uncanny valley. When people look at human replicas (pictures, statues, toys, robots), they have a generally positive reaction, which tends to get better as the replica gets more realistic. But as the replica becomes close to but not quite “perfect,” it falls into a distinct observed dip (or valley) in human perception, and can prompt actual revulsion. We see the effect in robotics and computer animation (it has been particularly noted in motion-capture films). The replica no longer looks like a replica; it looks like a real human, but with something wrong with it, something uncanny, and we react with apprehension. As Narrative Science starts giving Phillip K. Dick a run for his money and creates stronger, more graceful, even subtle prose—but without a perceptible human connection—reader reaction could actually become less positive.
I don’t know. Call it sour grapes again, but I do know that Spielberg’s The Adventures of Tintin creeped me out, and that the new more realistic voice automation I have to deal with when I call my bank irritates me a lot more than the old fashioned recordings used to. Let’s slightly paraphrase another fairly sensitive human writer: “O brave new world / that has such wonders in’t.” True that, but I still don’t know if I want to live in a bladerunner’s world.
Molly, Ramon – what wonderful replies! You can tell you guys are writers… Kundera and Shakespeare and Phillip K Dick – Oh my!
Ramon, you should have included the start of the line in the quote though “How beauteous mankind is!”
How beauteous indeed! Not that technology isn’t, of course—I just come down with you on the whole enhance/replace equation.
Now here’s a point that could add to either side of the issue: There are already computer programs that create art – or something like it.
I read about AARON a while ago, though I can’t find any websites about it that don’t look like they haven’t been updated in 15 years.
The story goes: Back in the 70′s, Harold Cohen began work on a computer that could create its own drawings. And so AARON was born. Harold Cohen wrote a number of essays about his efforts to improve the drawings that his program AARON produces, and one quote in particular from 1994 really jumped out at me – and suggests that some of the activities we’ve always known to require human creativity may not be as complicated as we think.
Still, it comes down to perception and subjectivity: this computer program creates drawings, but do they qualify as art? And if computers create content, does it qualify as real writing?
—–
From http://www.kurzweilcyberart.com/aaron/pdf/furtherexploits.pdf
AARON exists; it generates objects that hold their own more than adequately, in human
terms…and it does so with a
stylistic consistency that reveals an identity as clearly as any human artist’s does. It
does these things, moreover, without my own intervention. I do not believe that AARON
constitutes an existence proof of the power of machines to think, or to be creative, or to
be self-aware: or to display any of those attributes coined specifically to explain
something about ourselves. It constitutes an existence proof of the power of machines
to do some of the things we had assumed required thought, and which we still suppose
would require thought — and creativity, and self-awareness — of a human being.
If what AARON is making is not art, what is it exactly, and in what ways, other than its
origin, does it differ from the “real thing?” If it is not thinking, what exactly is it doing?
-Harold Cohen
—–
If you’re interested in reading more about AARON, I’d suggest checking out http://www.kurzweilcyberart.com/aaron/hi_essays.html
Very interesting Samantha! In my opinion, what AARON is doing is “practicing” – using examples of previous work to generate new work for the purpose of developing skills. Creating Art, in my opinion, takes more than that.
On this front, I found a very interesting article this weekend that speaks to this topic EXCELLENTLY. It is long, but well worth the read – describing attempts to get computers to generate language, how that linked to the Search industry (which ultimately moved away from writing code to try to evaluate language, to writing code to evaluate connections between sites), through to the metadata ontologies of online shopping sites and social media platform. It has a rather provocative title which I will include here – http://nplusonemag.com/the-stupidity-of-computers – but it is actually pretty optimistic about the future for computers and language/ Art. Here is a sample quote – The increasingly self-referential and allusive nature of art has already made “derivative” less of a pejorative, and the ability to mechanically process huge amounts of data with computer assistance will play a larger role in the construction of art of all kinds. David Shields in his book Reality Hunger was wrong to say that fiction as an art form is dying out; it just awaits new creators to reinvent the form for a more quantitative age.
Some people internally have referred to mahine translation in this context, so (to give the Machine’s some help in this debate) I thought I would include a link to a case study I worked on about that recently, about how recent improvements have made it viable in the Enterprise space here
Yebbut the rock paper scissors machine wins by…cheating. And also — could a machine figure out why, given two identical stories, one written by a machine and one by a human, I’d prefer to read the one written by the human?
I think we sometimes mix up the ‘whether or not’ question with the ‘why?’. You know — paintings by numbers can be pretty good these days…but most end up in Godwill.