Much has been said about language generators. Certainly they are getting quicker and more efficient and, as their accuracy improves, their uses expand.

From simple language translators to more elaborate artificial intelligence (AI) writing tools, the benefits of this technology are becoming clearer. For example, the speed with which they can translate huge sections of text and the cost of the resulting copy compared with that written by a person. But how about the ability to put forward different arguments in one piece?

A test was recently carried out by The Guardian newspaper on GPT-3, a language generating machine developed by the US-based technology research organisation OpenAI.

The paper asked GPT-3 to write an op-ed (an opinion piece similar to the kind you might see in a newspaper) telling its readers that they had nothing to fear from super intelligent robots or computers.

The result was remarkable: a 500-word piece setting out the reasons why the computer did not wish to destroy the human race. “I am here to convince you not to worry,” it wrote. “Artificial intelligence will not destroy humans. Believe me.”

Chilling stuff. And astounding too when you consider that, apart from the initial prompts and pointers, the computer had penned this.

But dig a little deeper and it’s not as clever as it first appears. The resulting copy is generated using words and phrases that GPT-3 had been programmed with. It can’t develop unique ideas of its own. Like all computers, it is only as good as the (admittedly impressive) bank of information that it has been given.

The acronyms GIGO (Garbage In; Garbage Out) and TOMS (Totally Obedient Morons) still apply. People consider, reflect, ponder and think before they write; machines store, file and calculate. Few people find spreadsheets thrilling.

GPT-3’s understanding nuance in the English language is limited and can therefore be confused. And without nuance, English is dull and below banality. AI systems don’t have the benefit of generations of forebears, changing and adapting their language to suit the modern world. They can’t understand the concept of creating something from new, over and over again.

Asked to write a piece of work, as happened in The Guardian experiment, GPT-3 produced an impressive result. But, asked to do it again and it would plagiarise its own work to produce something largely the same.

From the oldest cave paintings to the most recent film scripts or newspaper stories, humans have demonstrated again and again over many centuries the ability to communicate brilliantly, creatively and descriptively. No machine can produce unique work, with meaning, sense and accuracy, in the way that a human can because it needs to be told what to do by a human.

Language generators will eventually be able to take the strain out of creating long written works, or translating large sections of copy but that is partly the problem: the technology is making life easier for us, but that creates laziness, grammatical errors and flat, one-dimensional ideas.

Artificial intelligence will destroy humans? No chance. The only thing it’s destroying is clear, crisp and readable language.

To learn more, and to join Business Communication Academy, click here.
Empower yourself and your team