Aritifical Intelligence is all about speed. I don’t mean just the pace at which the technology trawls through vast databases of knowledge in order to generate a new combination of words, images, or sounds. I mean the speed at which the end user comes to expect results.
We’ve been living in a world of instant reactions for a long time. Email placed an unseen pressure on us to reply as soon as possible. The coming of social media meant that a piece of news (or indeed fake news) lost value if it was not reacted to immediately. Even words themselves (which, after all, take time to type) were always going to lose out to the quickfire response of an emoji: like, love, care, cry, an of course, hate.
A system I use to build many client websites, Divi, has just incorporated an AI engine into its software. It combines many of the tools people know from ChatGPT and MidJourney. At the moment it’s early days. But it will speed up, and ultimately may replace words as you type.
Imagine: immaculate prose at the touch of a button; the perfectly-suited image; the expertly-written bit of computer code. What could possibly go wrong?
Well, plenty, it seems. Just as in communications with humans, the answer comes down to context, audience and nuance. All of these are vulnerable to unguarded AI intervention.
Divi AI, for example, has a tool to change the tone of your writing into any of 18 appropriate moods: creative, professional, funny, serious, compelling, engaging, sarcastic, informative, academic, and so on. You can also switch between ‘beginner’ and ‘expert’.
At the moment, it’s quick but primitive. For example, ‘expert’ writing simply runs through a learnt model of synonyms, picking out those with the most academic-sounding Greek or Latin roots, and deposits the answer as if your writer has swallowed a thesaurus for breakfast. ‘Simple’ is mean to choose words that are less complex. But in practice it doesn’t work that way.
There’s a long way to go. But the critical thing is this. For the foreseeable future, human intervention is still required in deciding whether a critical message has hit the mark. That’s not just true of sensitive, legal and academic texts, but also of journalism, advertising and any kind of work that requires absolute accuracy and judgment.
I asked Divi AI to rewrite this last paragraph for younger readers. Here’s the response.
‘There’s still a long way to go, but here’s the cool thing – human intervention will still be needed for a while to figure out if a super important message has really hit the bullseye. This goes not only for serious, legal, and school stuff but also for news, ads, and anything that requires total accuracy and judgment.’
Hmm. I’m not sure I would swap ‘cool’ for ‘critical’, ‘or ‘school stuff’ for ‘academic’. And to check writing for this kind of thing takes time. In my judgement, no speed incentive is ever going to replace the absolute need to get it right. Maybe that’s why, for the moment, I’m still in a job that a computer has not replaced.