I found this article very disturbing.
Any of us could be algorithms - except eightbitsamurai. No, not 8bit, and I'm maybe 98% sure kleinbl00 is human. Mostly.
One of these baseball stories was written by a computer. One was written by a human. Can you tell which?
“Things looked bleak for the Angels when they trailed by two runs in the ninth inning, but Los Angeles recovered thanks to a key single from Vladimir Guerrero to pull out a 7-6 victory over the Boston Red Sox at Fenway Park on Sunday.”
“The University of Michigan baseball team used a four-run fifth inning to salvage the final game in its three-game weekend series with Iowa, winning 7-5 on Saturday afternoon (April 24) at the Wilpon Baseball Complex, home of historic Ray Fisher Stadium.”
The first one sounds spammy. Not mentioned by the article: most writing algorithms are also tuned (however mildly) for SEO since their whole purpose is to generate content for algorithmic search engine ranking.
In other words, algorithms write for other algorithms, not for you. See enough of their bullshit and you'll recognize it quickly.
That doesn't really matter, though. Both blocks are wretched. Copywriting is a fairly thankless task at best but a decent writer can make the boring stuff interesting. The top block is an artificial attempt to make statistics interesting while the bottom block is burying the lede. What the article proves is that humans can write as badly as algorithms, not that algorithms can write as well as humans.
Back in the dark ages (1991) we decided to type gibberish in 4-15 letter strings into MacWrite and then spellcheck it. The results sounded remarkably similar to Skinny Puppy lyrics. They were not, however, Skinny Puppy lyrics and while many people couldn't immediately tell which was computer and which was Nivek Ogre, one is definitely better.
A Pollack is still a Pollack even if most people just think it's drips on canvas.
- A Pollack is still a Pollack even if most people just think it's drips on canvas.
It's much the same for serialist music. when it's performed live and someone misses a note, it's surprisingly easy to notice.
- Back in the dark ages (1991) we decided to type gibberish in 4-15 letter strings into MacWrite and then spellcheck it. The results sounded remarkably similar to Skinny Puppy lyrics. They were not, however, Skinny Puppy lyrics and while many people couldn't immediately tell which was computer and which was Nivek Ogre, one is definitely better.
I'd be interested to see this if it was still hanging around the internet anywhere. would this even work with modern spellcheck?
I appreciate all the responses to this and have been thinking about my negative emotional reaction to the article .
It might be related to my need for trust. I'd like to think that a writer has thought about her text. Sure, dishing out statistics can be done quite accurately by an algorithm, but I'm somewhat horrified to think that I'm reading a novel or opinion piece generated by algorithms.
I hang up on robocalls, I delete robot-generated spam (which is most likely all spam). Irrationally perhaps, I want to know that a human consciousness, a sentient being, has written a personal essay or editorial. The possibility that everything I read has no human author creates distrust between me and texts --
The whole question of robot-authored texts somehow reminds me of 42 - the answer to the question of life, the universe, and everything.
Because a computer came up with the answer, it made no sense to the humans.
and yes, wasoxygen, Barthes says we should judge a text on its own merits regardless of how it was authored -- but that's if we are engaged in the text and want to judge it, read it, understand it, identify with it. We become engaged because we somehow trust that it has a message for us. To those who want messages from robots, that's fine. Robots might well have something to say.
I loved the SNL clip pseydtonne.
- I'm somewhat horrified to think that I'm reading a novel or opinion piece generated by algorithms.
But the algorithm was written by a human. So it's not robot-authored, so much as meta-authored by a human.
In software engineering, we have the concept of metaprogramming. Code which produces more code in a programming language. It's a great way to write many similar-but-not-identical blocks of code.
If I write an app which uses metaprogramming, no one would claim I didn't write the app, or that it's less of an app, or that I'm less of a programmer. Why is code which produces a natural language different?
What about algorithmically generated art?
What about digital art?
- but I'm somewhat horrified to think that I'm reading a novel or opinion piece generated by algorithms.
Many people already enjoy procedurally generated novels. Granted, it's not all there yet, but the idea is neat. Most story tropes are formulaic, which makes it easy to generate an outline. Depending on the subject you want either technical or flowery language. Or perhaps plain. All of these are fairly standard stuff.
The hard part is to introduce new ideas. Which, at the moment, is only possible by a human. The cool thing about computer written stories (or articles) is that they can be generated on the fly, and have user interaction.
Certainly a human writer could do this, but not for thousands of readers. And not nearly as quickly.
- I hang up on robocalls
I hang up on human calls who read scripts as well. Hell, most of the time I prefer to deal with a computer, because they act in a predictable way.
- I delete robot-generated spam (which is most likely all spam)
Nope. A lot of 'spam' is written by humans as well.
- I want to know that a human consciousness, a sentient being, has written a personal essay or editorial.
How are you certain that I'm a 'human consciousness or a sentient being'? What if your the only sentient being and we are all philosophical zombies? Would that influence the way you interact with people? What happens when we get to the singularity and migrate our brains over to computers? Will you discriminate against transhumans?
How do you accurately distinguish whether something has a consciousness? Are you going to discriminate simply because someone acts in a more standardized (rather than a chaotic/'human') way?
- The possibility that everything I read has no human author creates distrust between me and texts --
If anything, it gives me more trust. A computer only does what it's told (so far). It has no ill-intent unless the creator gave it one. While a human is almost certainly out for their own benefit.
- Because a computer came up with the answer, it made no sense to the humans. Maybe there's no connection.
While '42' was meant to be a hilarious example, I think that the idea has merit. I'd personally say '3' is the answer, not 42. But as far as a computer determining the 'answer to life', that's kind of silly. It'd most likely just ask you to clarify. What answer are you looking for? The non-sensical answer was in response to the non-sensical question. Much like if you try to google 'not elephants' you'll almost certainly get elephants.
- Barthes says we should judge a text on its own merits regardless of how it was authored
I would agree. In fact, I'd agree that's the point of language: to communicate ideas and to obtain information and other perspectives. In which case the author/speaker has no effect on the reception of the words.
- We become engaged because we somehow trust that it has a message for us.
I don't. I don't trust anything that doesn't come from my own mind. When I began responding to this comment (like I do all comments), I set out to respond with my thoughts. I don't care whether there's an overarching message or idea behind your comment. Only that it is there to comment on and that each sentence and the body as a whole interests me. You could be a computer, a human, a dog, a cat, etc. It doesn't matter.
- To those who want messages from robots, that's fine. Robots might well have something to say.
Well it depends on your goal and what you are looking for. Sports facts and news aggregation? I have that automated and sorted automatically. Both reddit and hubski do that fantastically. And hell, not even all the posts are user submitted, some are by 'bots'.
As far as robots having something to say, I'd say they most likely will, once we reach a certain point. But right now? They are just pieces of code that are ran.
But you could easily say the same about humans. Humans work off biological scripts. Just complex ones.
Finally, I'd like to leave off with a clip about conversations. It's part of one of my favorite shows, and brings up a good point.
And I'd like to point you over to a fantastic series (which has been merged into a movie) called "Time of Eve" which covers the robots/humans dynamic. It takes place in the future where robots have become indistinguishable (visually and mentally) from humans. But robots are still 'slaves' seeing as that's how technology is now. To distinguish, the robots are forced to have a ring above their head, to signify their place (you could almost relate this to skin color back in the slavery days). However, in the show, there's a certain cafe in which the rule is to not discriminate between robots and humans. And as such, the robots turn off their ring. The result is that you can't distinguish between the two groups, besides some 'obvious' cases (humans talking about robots like others, some robots have some faulty parts, there's a robot that clearly looks like a robot, etc).
It's a fascinating watch, and really covers this kind of topic brought to an extreme. Would you be uncomfortable treating a robot as an equal? Why or why not? What is it that humans have that an indistinguishable robot doesn't? It quickly turns into 'Well humans are humans, and robots... aren't' discrimination.
- How are you certain that I'm a 'human consciousness or a sentient being'? What if your the only sentient being and we are all philosophical zombies?
Even conversations on IRC seem to be generated by humans.
Kyon and the Talking Cat is a great clip and I totally agree with the cat. What I am saying (or writing) might well be understood in a different way by the other. It might even be heard as gibberish.
- Would that influence the way you interact with people? What happens when we get to the singularity and migrate our brains over to computers? Will you discriminate against transhumans?
Re transhumans: If a robot is conscious of inequality and expresses that awareness, I would certainly pause. If a robot is programmed to get revenge when it is treated badly, our engineers might have some 'splaining to do.
- You could have deliberately made a spelling error to trick me into thinking you were human. Still, I'd like to think that hubski is perhaps 99.9% human. -
There's still the philosophical zombie issue. And the identical robot thing (which obviously isn't here yet). But yes, you can be safe in assuming 99.9% of hubski (and even reddit) is a biological organism that we call humans.
- that's one reason why I love meetups or reports of meetups. I've seen some faces.
See 'Time of Eve'. Granted, that's not here yet. So meeting in person is a safe way (for now) to determine whether someone is a biological organism.
- I've received postcards written in ink by what seems to be hands (although I can't be sure).
You can make handwriting fonts and print them. No guarantees post cards are written by hand. IIRC there's also handwriting services (like in the movie 'Her').
- Even conversations on IRC seem to be generated by humans.
Correct. "Chat bots" aren't quite there yet either. Though some (like cleverbot) have passed preliminary turing tests.
- Kyon and the Talking Cat is a great clip and I totally agree with the cat. What I am saying (or writing) might well be understood in a different way by the other. It might even be heard as gibberish.
That's why I linked it. It's sort of the same idea. Even other humans can write something which is completely different than how you understand it. So at what point is a 'connection' really made?
- I leave that to future generations - although the piece I posted recently about head transplants moves in that direction.
It's rapidly becoming a current problem. I'd say at least within 100 years. The singularity is predicted to happen within current young adults' lifespans. But I bring it up because it's an increasingly problematic thing. Vegans/Vegetarians continually bring it up as a moral basis for their dietary preferences. If animals are treated with the same conscious respect as humans, why not machines?
- Re transhumans: If a robot is conscious of inequality and expresses that awareness, I would certainly pause. If a robot is programmed to get revenge when it is treated badly, our engineers might have some 'splaining to do.
The unfortunate fact is that early consciousness can't be aware of inequality. Certainly at some level that is obtained though. And as for it being 'intentionally programmed', that may very well not be the case. When it comes to computer vision, we don't actually program how to recognize things. The computer 'learns' that on it's own. Once the code is sufficiently advanced, certainly the computer could learn to recognize inequality?
To give a more retro example, there's a game called 'creatures', in which you take care of a virtual creature. This creature is unlike many you are familiar with. it has a dietary track, in-depth health meters, and a brain that is composed of many parts, like a human. It has a visual part, where it can learn to recognize objects. A speech part, in which it learns to identify and be able to speak and respond to speech (through text input/display on the screen). It can learn routines, and how to fend for itself (though at the beginning you must teach it what's good/bad). And it can feel pain/pleasure. Hell, it can even fall to skinner boxes. These creatures can mate and have offspring. Much like a real creature. This creature can feel regret and have "guilty pleasures", in which it likes something and will do it, even if punished.
Is this creature worth respect? Is it worth teaching things? Certainly it's rudimentary (as it's a commercially available old piece of software). But if something that old can be that complex, what's stopping it from happening in the future?
Along the same lines, transhumans are literally just a human mind uploaded into a machine. Are these people no longer deserving of respect simply because they chose a different lifestyle? How do you distinguish between a transhuman and a full robot?
As I said, go watch 'Time of Eve'. it really nails down these points and is one of my favorite movies/shows.
- So at what point is a 'connection' really made?
Humans seem to have capacity to create profound connections with inanimate objects, to even turn these objects into companions (think "Watson" and Tom Hanks in Cast Away)
At what point is a connection "really" made? I guess it all depends on what you think "really" means. I take "connection" to mean a sense of a shared and collaborative reality.
If you mean human to human, the point when you feel connected can be immediate - first glance - or it can be gradual through shared experiences.
Our connections can give us a sense of being useful, valued, cherished, loved. Our connections help us create our identity and sense of who we are.
Luckily, there are many ways of connecting to other humans. Like Hubski.
Ah, love, let us be true
To one another! for the world, which seems
To lie before us like a land of dreams,
So various, so beautiful, so new,
Hath really neither joy, nor love, nor light,
Nor certitude, nor peace, nor help for pain;
And we are here as on a darkling plain
Swept with confused alarms of struggle and flight,
Where ignorant armies clash by night.John Henry is an improbable folk hero; he came to a bad end after opposing an advance in technology that made life better.
Telescopes reduced God's dominion over the cosmic spheres, washing machines reduced the household's dominion over women, automobiles reduced the dominion of horse manure over streets, Chinook and Deep Blue reduced the dominion of experts over board games.
At every step people worried what these advances meant, while at the same time embracing and adapting to them.
Ever since Barthes told us the author is dead, we have been expected to judge a text on its own merits, without worrying about the author's intentions, culture, psychology or (I suppose) whether it is made out of meat. Perhaps that is a good way to look at computer-generated writing.
These baseball blurbs both sound pretty crappy to me, desperately trying to make a few statistics read like a story. If it's any comfort, the algorithm almost certainly has a phrasebook based on the many ways human sportswriters have devised to express baseball concepts and it simply mixes and matches these terms with less creativity than it takes to play Mad Libs.
Nevertheless, it seems inevitable that a computer-generated novel will one day become a bestseller.
If algorithmic writing creeps you out, Der Bibelschreiber will surely give you the willies.
For sports and market stats, an algorithm could be an improvement. In those cases, flair or the author's voice often isn't what you are after.
I'm not much bothered by this, actually. If a computer can write a better novel, then I will probably start making friends with computers.
I agree about the sports and stock stuff. After all, the National Weather Service has been doing this with forecast data for twenty years or so.
Just keep it collegiate. Otherwise...
A bit irrelevant but this reminds me of a 'idea generator' I made a while back
http://pastebin.com/0nx2kq4y I know it's nothing in comparision but it was really easy to make, I had a set of templates which had text with some keywords which inserted/removed more text. For example "Write a $bookViewpoint $booktype about a $character as $gender battles through $location to $location2 where $gender finds $character2 under the influence of @drug"
becomes
- Write a second person novel about an orphaned Scientology messiah as he battles through The Great Forest to Solitude where he finds The Real Slim Shady, under the influence of a alcoholic screen cleaning wipe
- One of these baseball stories was written by a computer. One was written by a human. Can you tell which?
“Things looked bleak for the Angels when they trailed by two runs in the ninth inning, but Los Angeles recovered thanks to a key single from Vladimir Guerrero to pull out a 7-6 victory over the Boston Red Sox at Fenway Park on Sunday.”
“The University of Michigan baseball team used a four-run fifth inning to salvage the final game in its three-game weekend series with Iowa, winning 7-5 on Saturday afternoon (April 24) at the Wilpon Baseball Complex, home of historic Ray Fisher Stadium.”
This was pretty easy for me, it's hard to describe why but maybe after a while of going through that pastebin you'll know what I mean.
The difference between a machine and a human is pretty clear; human writing has a tone.
Both of those samples are nearly identical because it's technical facts. Why bother with technical facts anyway? I certainly don't want to read a long article on something I could just read wikipedia for. I read stuff written by people because I want their opinion. And I try to reflect that through my writing as well.
But this ties into something I've been thinking about a lot. Why the discrimination? If a robot can write as well as a human, why be picky about who writes? You don't say "Well women write worse, so we'll let them write because they need jobs too."
This is the thing with the computer-generated novel. To what extent is the computer actually "writing" that book? I mean how much human assistance is it being given, in terms of instructions and most specifically, pre-generated phrases and pre-linked words, to create it?
Eg the 50shades generator (of which there are several): http://www.xwray.com/fiftyshades/
It's being loaded up with so many pre-written phrases that the computer not really "writing", it's mixing and matching.
I'm struggling to articulate this, but what I currently perceive is that for any credible "computer generated" literary work, as opposed to eg a match or weather report, you need so much creative human input that the computer isn't actually having to do very much.
I have wondered about getting computers to write sex scenes. In terms of pornographic sex. where there is limited interest in emotional development, and one is really looking at a sequence of actions punctuated with a few "oh baby" and "slam it in" etc, I think a machine could do that brilliantly. I actually think if you wrote the right formula to start with, just changing a few variables each time you could mix and match and generate reams and reams of erotica, and all you would have to do is edit it a bit, set the scene at the start, and probably make the last paragraph seem more original.
Well arguably, at what point does the computer have ownership over the work done? Certainly there are generators which simply play mad-libs. And then there's stuff that fluctuates grammar as well.
But does the computer own the writing generated? Or did the human write it through proxy, by simply writing code to generate the writing for him?
No matter how complex the algorithm, a human certainly wrote it. So depending on what you mean, the entire work (even these 'roboarticles') could be claimed by a human. if you generate HTML code, is that the computer writing the code? Or is it the human writing code to generate code on his behalf?
Taking that into thought, and removing the "object ownership" idea, your parents "generated" you, so does that mean everything you write is "written" by them, via proxy of creation? That is, they worked to create and raise you in a certain way, at which point you'd output the desired text. Just like the programmer, they can't be sure of the exact output, but are responsible by proxy. Does that mean your work is yours? What does ownership even mean?
In the case of mad-libs generators, I'd say that the final product is written by the computer, but the preset text is not. That is, cleverbot is communicating through lines of chat that has been given to it. The chat log is written by cleverbot, even though the individual lines were outsourced to humans.
At what point does the computer obtain autonomy? At what point do humans obtain autonomy? Did you really write the book? Or did the computer write the book in response to your key presses. Did you really write the book? Or did the pencil you manipulated do so?
- you need so much creative human input that the computer isn't actually having to do very much.
I wouldn't say so. Perhaps to get the specific grammatical structure correct, sure. But I think after a point, you've written a program that can write on it's own, without relying on humans.
Again, it boils down to, at what point do you distinguish a program from it's creator?