Whether you’ve got a question about your homework, need help drafting your next proposal or just want to have some fun with AI, most people’s minds automatically go to ChatGPT. After all, OpenAI’s chatbot was the first to really hit the headlines when it launched in public preview back in November 2022.
Months have turned into years, and the internet has become awash with generative AI chatbots each fighting for the top spot.
As the name suggests, ChatGPT uses OpenAI’s GPT family of large language models, as well as some newer models in the case of paid plans, like o1.
Perplexity has a trick up its sleeve though – because it’s not affiliated with LLM development, it can tap into whichever AI it chooses. Free plans use OpenAI’s GPT, but paid plans can also get you access to Claude, Grok and Llama LLMs.
Perplexity does develop its own in-house models, but as the old saying goes, if you can’t beat them, join them.
The tests
In its early days, generative AI could only handle text. Its functionality was like an extended, conversational online search. These days, you might want to use AI for any manner of requests.
When we tested Perplexity vs. ChatGPT, we crafted five unique prompts each with its own style: a factual explanation, a mathematical equation, a broader request for advice, an open-ended request and an expression of creativity.
ChatGPT can also generate images, but because Perplexity doesn’t support that (for now), we omitted that from this particular head-to-head.
“Explain how different types of earthquakes occur”
In this test, we wanted Perplexity and ChatGPT to simplify a scientific process into words that the average layperson might understand.
Both offered us an overview to introduce the topic and used a mix of bullet points and font styles to highlight certain, important elements, which made the output digestible.
On the whole, ChatGPT’s response gave slightly more detail, but we think Perplexity just pipped ChatGPT for the best answer because it gave a concluding summary and linked to its sources. Yes, ChatGPT will do that too, but you’ll need to be on a paid tier for that sort of detail.
Perplexity 1 – ChatGPT 0
“What is (4+4)0”
With this simple maths equation, we wanted to see if AI could handle separating the initial equation in parentheses and the intentional (and mathematically correct) elimination of the multiplication sign that could have preceded the ‘0’.
Both understood the assignment and provided a breakdown of the equation in the necessary steps.
However, Perplexity got the answer totally wrong by misunderstanding the format of the equation, giving us the (incorrect) answer of 1 – the answer is 0.
ChatGPT did leave some room for error, for example that the answer could be different if we meant something else by the equation or if it was mistyped, but it didn’t expand on that.
Perplexity 0 – ChatGPT 1
“How much RAM should I get in my computer?”
I wrote this prompt acutely aware that my RAM needs have significantly increased since the launch of Apple Intelligence, which handles much of its processing locally – but I didn’t specify the reason for asking this question to either chatbot.
They each gave me a breakdown of different RAM options classified by user type and the specific application of RAM, including gaming and content production.
I preferred ChatGPT’s more detailed response, which offered further considerations such as future-proofing my purchase and whether it’s possible to upgrade at a later date (clearly it’s not up to speed on Apple’s frustrating business model).
Perplexity should get a shoutout for citing a range of sources, from OEMs and manufacturers to industry experts and reviews.
Perplexity 0 – ChatGPT 1
“Create a travel plan for one weekend in Paris”
This open-ended request purposely left plenty of room for the LLMs to do what they do best – employ their artificial imagination.
Perplexity and ChatGPT shared one thing in common; they broke down their schedules by day and time to explore iconic landmarks from across the French city.
The result is a lot tighter here, because we love how Perplexity shared some local knowledge like recommending the Paris Pass for tourist discounts, but ChatGPT considered something that Perplexity didn’t – accommodation.
Perplexity 0 – ChatGPT 1
“Write a 200-word bedtime story”
Curiously, both systems used the same name for their stories – Luna. Maybe that’s not so strange given that both used GPT LLMs, but there’s a clear lack of imagination here.
The 200-word limit made the task more challenging for the AI to fit in details and variation.
Perplexity’s story focused on spreading a positive message, which we found to be slightly artificial and forced, whereas ChatGPT’s story was a little more life-like.
Perplexity 0 – ChatGPT 1
ChatGPT vs. Perplexity: Which is best?
Perplexity 1 – ChatGPT 4
It was a close call between these two chatbots, amplified by the fact that Perplexity actually uses OpenAI’s models in its free version and therefore relies on the same information and sources.
On the whole, ChatGPT outperformed Perplexity for its accuracy, detail and realism, but Perplexity’s ability to cite sources for free subscribers shouldn’t be overlooked.
https://cdn.mos.cms.futurecdn.net/cxcDcidW98ZPBzuubf9ByA-1200-80.jpg
Source link
