Perhaps — with some caveats.
Considering the rapid pace of innovation in communications broadly, the email sector has remained surprisingly little-altered over the past 25 years. But maybe that’s about to change.
We’re now seeing developers building tools that can create email copy using text-generation technology. This past June, artificial intelligence research lab OpenAI unveiled a language model known as GPT-3, which has quickly led to several automated writing tools being developed.
Just what is GPT-3? Here’s a definition according to The Great Book of Wikipedia:
“Generative Pre-trained Transformer 3 is an autoregressive language model that uses deep learning to produce human-like text. It is the third-generation language prediction model in the GPT-n series created by OpenAI, a for-profit San Francisco-based artificial intelligence research laboratory.”
In a nutshell, automated writing tools built on GPT-3 send bits of keyword text provided by an author – otherwise known as “prompts – to OpenAI’s cloud service, which instantaneously sends back full-flowing text that’s deemed appropriate and accurate based on the statistical patterns it recognizes in the online text.
Even though GPT-3 technology accesses a vast information bank of training data comprising nearly 500 billion tokens in cyberspace to “derive” the copy, there’s always the possibility that the results could end up like the early attempts at automated language translation at the start of the 21st century – garbled and awkward. However, with more AI “practice” and crowdsourced feedback, we’ve seen an established service like Google Translate deliver excellent translations for most commonly used languages like German, French, Spanish, Chinese and Japanese.
Languages such as Hungarian, Turkish and Lithuanian are another matter – presumably with more seasoning needed to get those more esoteric tongues in ship-shape for AI translation.
Facebook, which has developed its own “walled garden” automated translation app, appears to be lagging Google considerably in the quality of its output – even when working in the most common languages like translating from French to English.
For now, the most practical applications of the GPT-3 language model look to be in the realm of business email writing, rather than for long-form business thought-pieces or most forms of creative writing. In email communications, the author can jot down three or four key points and let the writing application do the rest. In this manner, instead of having to craft a memo completely from scratch, authors can provide key snippets — then take a moment or two to edit the proffered text before sending the email on to its intended recipients.
For those of us who write for a living, such a procedure might not seem particularly attractive. But for the many people who dislike the task of writing business communications — or find it laborious and too time-consuming — the new AI-powered writing may well be a welcome tool.
OpenAI’s automated writing service is on the pricey side today, but we can expect that it won’t take long before costs borne by end-users start to drop precipitously — no doubt due to the proliferation of free services subsidized by the same monetization model that now supports Google Maps and Google Translate.
And this brings up a question that people should start to think about sooner rather than later: Who will own the copyrights to the automated texts generated in this manner?
It’s only a matter of time before the leading purveyors seek to leverage their ownership of the freeware and the licenses they grant to use it – thereby giving them the ability to promote or censor whatever information they please.
The social acceptability of this medium could also be eroded when the volume of ghostwritten email masquerading as personalized communications begins to overwhelm people’s inboxes. At some point, email recipients will come to realize that any message that doesn’t include a disclaimer such as “I am the author; please disregard all spelling and/or grammatical errors,” can be marked as spam and routed automatically to the recipient’s junk email folder. In such an environment where we’ll have a perceived quality demarcation between “real” and “manufactured” writing, we may find ourselves in the same place as we are today with tweets — that is, weighing if they are the work of humans or bots and judging their worth accordingly.
In other words, the new “next thing” in email communications won’t be happening without its share of issues and controversies – along with more than a little disruption. It will be quite interesting to see how it all unfolds in the coming years.
What are your thoughts on the role of AI in writing? Is the technology poised to become mainstream quickly, or will it remain more of a curiosity for a good while longer? Please share your thoughts with other readers.
One thought on “Is automated copywriting the next big innovation in email?”
AI has a long way to go with email prompts. If you’re someone who can’t spell or line up singulars and plurals properly, it will set you on an improved grammatical track. But if you’re a writer interested in style, you will find that Google or other services have a surprisingly small vocabulary and will force you into generic writing.
That’s perhaps OK for business emails, but not very useful for me as a music critic, for example. In the course of writing about music originating in foreign countries, I have frequent occasion to use Google Translate. It usually produces a decipherable result in romance languages. But German word order, with verbs at the end, flummoxes it to the point of hilarious incomprehensibility. (A German friend tells me the algorithm is less grotesque from English to German.)
Still, how can one complain of such instantaneous, nearly bionic assistance? I stop to think: How many shelves of books or visits to a library would it take to accomplish what I have managed in a moment with an extra screen or two?
And this is just the beginning.