The chief executive of one of London’s leading artificial intelligence companies has warned his staff against “AI-sloppification” after an increase in documents written by large language models.

Victor Riparbelli, who co-founded Synthesia in 2017, told his staff that “no one wants to read” overwritten documents with “pages of fluff”, and that it was a “net-negative” for productivity. 

“Now that we’re all using AI to write, I’m seeing an increase in overly verbose documents,” Riparbelli wrote in a company-wide memo. 

“Large language models are great tools for pushing our thinking and helping us write faster but it’s a fine balance between writing with enough depth to convey complex problems and driving clarity by being concrete and concise. 

“When we write too long and verbose docs we place the burden of distilling the content onto the readers — so instead of one person spending ten minutes on sharpening a doc we have ten people spending ten additional minutes reading slop, leading to net-negative productivity.”

Synthesia uses AI to make realistic videos for employee training and marketing. The company sells to 90 per cent of the Fortune 100 — the biggest businesses in the US by revenue — and was valued at $4 billion at its most recent funding round in January. 

Riparbelli’s frustration with the overuse of large language models reflects the increasing quantity of low-quality content produced using AI tools, often known as slop. Since the launch of ChatGPT in November 2022, AI-generated content has become increasingly prevalent online and in the workplace. 

Business newsletter

The business editor’s exclusive analysis of all the latest financial and economic news.

Sign up with one click

Researchers at Stanford, Imperial and the Internet Archive found that a third of websites created since 2022 have been AI-generated, while a study by Originality.ai estimated that 54 per cent of posts on LinkedIn were believed to have been written by AI.

Riparbelli encouraged his staff to continue experimenting with large language models but said they should “make us more concise, not more verbose”.

He said that the ever-increasing wave of “AI-sloppification” would put a premium on meaningful communication. “AI will make great human communicators even more valuable than before,” he said.