Another Type of Slop: LLM Augmented with Biased Data

Perplexity.ai ranking a product as #1 because the product has a webpage claiming they're #1.

Recently I've been using Perplexity a lot to do simple research, which has saved me a lot of time reading through reddit comments and individual web pages.

Many of my Perplexity runs give good results, and some even find sources that I won't be able to find normally, since it generate different search engine queries than the ones I would use.

While experimenting with the tool, I remain cautious of the responses it provides and always check the linked sources when I feel something is not right.

Below is an example of Perplexity using biased sources and generating a highly misleading answer:

Ignore my grammar mistake. Response looks convincing at first glance, since it has referred to multiple sources

However, upon examing the sources, the claim that stocktotrade is the #1 is made by itself, and I wouldn't assign too much credibility to that.

The neologism slop refers to overly-hallucinated content generated by generative AI, and I think this example also counts. It's not hallucinating too much, but it failed to examine if sources (the truth based on which the response was generated) are trust worthy, and the response couldn't be replied upon either. Kinda like supply chain attack.