Elijah Clark: A Review of Perplexity AI, Rewritten by Perplexity Itself

In the ever-evolving world of artificial intelligence, Perplexity AI has emerged as a promising contender, offering a fresh take on information search and synthesis. Thanks to my school's collective effort in signing up, everyone at my university won a year's worth of free access to Perplexity Pro—a delightful irony considering our institution has labeled it an unreliable source.

After some hands-on experience, I've discovered both its shining strengths and a few quirks that might make you raise an eyebrow.

A citation sensation...

One of Perplexity's standout features is its dedication to citations. Unlike ChatGPT, which often feels like a magic trick where the rabbit appears without explanation, Perplexity lays its cards on the table. It diligently cites sources, making it less prone to hallucinations—a common ailment in the AI world where models confidently spout inaccuracies.

For instance, when I asked about section 179 tax code exclusions for hefty SUVs, Perplexity not only delivered nuanced answers but also detailed its search process, reassuring me of its thoroughness.

This transparency extends to its Discover tab, which summarizes sports, entertainment, and tech news with aplomb. While it doesn't dabble in political or international news (perhaps wisely avoiding potential minefields), it crafts engaging articles that could give your favorite news outlet a run for its money. Imagine NPR or BBC News Briefings delivered by an AI with a penchant for clarity.

...with contextual conundrum

However, not all is rosy in the land of Perplexity. The AI struggles with contextualism—remembering past interactions and building upon them. This is particularly evident in its voice assistant feature, which often cuts me off mid-sentence and loses the thread of our conversation faster than you can say "Siri."

It's a bit like trying to have a deep conversation with someone who keeps forgetting what you were talking about.

While Perplexity's voice assistants sound natural and pleasant (a definite plus), it falls short in delivering detailed responses verbally. Despite being better than other generative AIs in terms of voice quality, it lacks contextual awareness.

This is especially noticeable when compared to Apple's Siri—which, while not a generative AI, is better at picking up conversational cues and building on past requests or questions.

A mixed bag on source quality

Another area where Perplexity could improve is in differentiating the weight of its sources. I've noticed that it often gives equal weight to citations from Reddit as it does from Forbes or The Economist. This approach can lead to inconsistencies in the quality of information provided.

Much as Wikipedia serves as a valuable starting point due to its thorough citations, Perplexity’s strength lies in its citations – but will only ever be as good as its citations. I'd love to see Perplexity offer users more control over which sources it prioritizes.

On the brighter side, Perplexity Pro offers customization options that are strong offerings. With access to models like Claude 3 and GPT-4o, users can tailor their experience to suit specific tasks. These different models, including Perplexity’s own proprietary model, are each strong in different areas. I’ve noticed some models are better than others at creating images.

I've been experimenting with the "spaces" feature to create a specialized model for rewriting emails—complete with instructions on tone, phrasing, and sign-offs. It's like having a personal assistant who knows exactly how you like your digital correspondence served.

Perplexity AI is a robust tool with a knack for providing nuanced answers backed by solid citations. Its strengths in transparency and customization make it a valuable asset for those seeking reliable information. However, it struggles with contextual understanding.

But until it improves, it's still worth keeping an eye on this promising new player in the AI arena.

You can read the whole thread leading to the creation of this Expert Opinion.

Elijah Clark is a linguistics student who brings a unique perspective to issues of digital communication, internet accessibility, and the impact of broadband on diverse language communities. His focus is on issues in translation and variation within language, and hopes to go to law school and to advocate for low-resource languages and vernaculars near and far. This Expert Opinion is exclusive to Broadband Breakfast.

Broadband Breakfast accepts commentary from informed observers of the broadband scene. Please send pieces to commentary@breakfast.media. The views expressed in Expert Opinion pieces do not necessarily reflect the views of Broadband Breakfast and Breakfast Media LLC.