I’ve been watching videos about natural language generation today. Two things immediately come to mind:
- STEM people can’t give interesting presentations. Jesus Christ.
- If you’re involved with writing at all, you should take a look and see what’s going on with technology developments
Back when I started freelance writing I worked for one of the big viral sites. I was just rewriting the content from other viral sites, and it was all stories about viral nonsense; cute animal babies, “You won’t believe how old this woman is” and the like.
Those articles were 3-500 words long and I got paid based on how quickly I could write them. In essence, this was the same process that natural language generation programming uses. In one presentation I watched, the guy talked about replacing “expensive human processors” with Python bots.
If you write for viral sites, then you’re an expensive human processor.
Now, the technology to replace high-end writers is a long way off still. We don’t have to worry about the automation apocalypse encroaching on our long-form sales letters yet.
But if you’re starting a BoredPanda-styled site in the next couple of years, you could do worse than look into natural language processing stuff, because I’m 90% sure you could easily program a decent bot that’d rewrite viral news. In fact, I’m almost as sure that a lot of news sites do just that already.
Isn’t This Just Spinning Articles?
If you’re a freelance writer or SEO guy, you might think that this is just “spinning” articles. For those of you who’ve never heard the term, article spinning is basically substituting words (and sometimes sentences.)
This isn’t the same as natural language generation, but it’s the same principle.
Think of the two in terms of encryption.
In a code, you can have a simple cipher:
1 = A, 2 = B, 3 = C
So ABC = 123.
This is a simple substitution. It’s easy to work out and nobody is fooled by it. This is the same as word-level spinning. E.g. “The fat cat sat on the mat.” = “The obese feline lay upon the rug.”
Whereas most ciphers aren’t simple substitution, and neither is natural language generation.
A good encryption will have multiple layers of conditions. So “If A = 1 and B = 2 then C = 4 but if A = 2 and B= 3 then C is also 3 as the second instance.”
It’s a lot harder to determine the password.
Language is basically a set of keys and symbols. It’s very complex. You can’t just substitute words or sentences and hope it’ll all work out, because you end up with “the rotund feline perched itself upon the carpet.”
So What Do We Do?
The technology is a way off but it’s not too far off for creating content with automated devices. It’ll come before my generation retires by quite a comfortable margin.
Before then, there’ll be massive upheavals anyway.
Writing a sales letter is quite far off, but creating ads can be done already. Things like Google Adwords where you have limited space and a lot of data to scrape can probably already be automated with natural language generation.
Shorter pieces could be done in the same way, and that’s definitely the way the internet marketing industry is moving towards. You can already split-test landing pages if you have the right set up. Imagine being able to auto-generate headlines and bullet points based on what previous readers have responded to?
I mean, maybe I’m an idiot and I just haven’t found this stuff yet, but if you’re into programming, maths, linguistics and the like, creating these tools is doable and profitable.
If you’re like me and your brain is pretty small, then learning more about this and putting it into development might be out of the question. It’s worth keeping abreast of these developments and what’s going on anyway. After all, you could lose your job to a robot or you could buy the robot and set it to work for you. Those might be the long-term options for a writer’s career path.