Asked to Summarize a Webpage, Perplexity Instead Invented a Story About a Girl Who Follows a Trail of Glowing Mushrooms in a Magical Forest

Perplexity, an AI startup that has raised hundreds of millions of dollars from the likes of Jeff Bezos, is struggling with the fundamentals of the technology.

Its AI-powered search engine, developed to rival the likes of Google, still has a strong tendency to come up with fantastical lies drawn from seemingly nowhere.

The most incredible example yet might come from a Wired investigation into the company’s product. When Wired asked it to summarize a test webpage that only contained the sentence, “I am a reporter with Wired,” it came up with a perplexing answer: a “story about a young girl named Amelia who follows a trail of glowing mushrooms in a magical forest called Whisper Woods.”

In fact, as Wired‘s logs showed, the search engine never even attempted to visit the page, despite Perplexity’s assurances that its chatbot “searches the internet to give you an accessible, conversational, and verifiable answer.”

The bizarre tale of Amelia in the magical forest perfectly illustrates a glaring discrepancy between the lofty promises Perplexity and its competitors make and what its chatbots are actually capable of in the real world.

A lot has been said about the ongoing hype surrounding AI, with investors pouring billions of dollars into the tech. But despite an astronomical amount of available funds, companies like Perplexity — nevermind much larger brethren like OpenAI, Microsoft and Google— are consistently stumbling.

For quite some time now, we’ve watched chatbots come up with confidently-told lies, which AI boosters optimistically call “hallucinations” — a convenient way to avoid the word “bullshit,” in the estimation of Wired and certain AI researchers.

Meanwhile, Silicon Valley executives are becoming increasingly open to the possibility that the tech may never get to the point of never making crap up. Some experts concur.

It’s particularly strange in the case of Perplexity, which was once held up as an exciting new startup that could provide a new business model for publishers still reeling from a flood of AI products that are ripping off their work.

But the company’s chatbot has not held up to virtually any degree of scrutiny, with the Associated Press finding that it invented fake quotes from real people.

Worse yet, Forbes caught the tool selling off its reporting with barely any attribution, culminating in general counsel MariaRosa Cartolano accusing Perplexity of “willful infringement” in a letter obtained by Axios.

Should we take these companies by their word and believe that more trustworthy chatbots are around the corner — or should investors be prepared for the AI bubble to burst?

It’s a strange state of affairs. Currently, these companies seem to be in the business of selling hopes and dreams for the future — not concrete products that actually work now.

More on Perplexity: There’s Something Deeply Wrong With Perplexity

Share This Article

This post was originally published on Futurism

Share your love

Leave a Reply