It was about time. It is not the first conversational AI to be connected to the Internet, but it is the first service that offers results on a par with ChatGPT. Perplexity AI has been updated in a big way and the result is extraordinary. It is difficult for this small company to be the “Google Killer”, but it does confirm that the ubiquitous search engine needs to give it a twist to keep up with the competition that is coming.
The easiest thing is to try it yourself. Especially now that for the moment the tool has no traffic problems and works quickly. Perplexity AI is a conversational search engine, where we can ask him anything and he will answer us in the best way he can. But most importantly, unlike ChatGPT, this time we have an AI connected to the internet that offers results in real time.
A quick test is to ask about an event that just happened. In this case, the resignation of the CEO of Netflix, which we have recently communicated at Xataka. Perplexity AI correctly responds that Reed Hastings has resigned and that Ted Sarandos and Greg Peters are the new co-CEOs.
Perplexity AI is connected to the Internet, which allows it to provide updated answers in real time.
The interface is very similar to Google, with a central bar where to write the question. In this sense, from Perplexity AI they have understood very well what is the strong point of ChatGPT. It’s not just how accurate it is in responding, it’s also how easy it is to use.
The importance of marking sources
In the case of the Netflix question, the sources from which it gets updated information are marked just below the answer. There is no objection here, since it is based on Netflix itself, on Wikipedia and on two related recognized media such as CNBC and the Hollywood Reporter. There will be specific cases where the information is not correct, but by clearly linking the sources we can verify it for ourselves.
The reliability of the answers is not entirely accurate, but we are linked to the different sources from which the data is obtained.
The source reference is precisely one of the most sensitive aspects in a search engine. As with ChatGPT, the search engine works equally in Spanish if we make the answers in our language. It is interesting that when changing languages, you logically opt for local sources.
If we ask you about the price of electricity in Spain for tomorrow, you understand the day we are on. The answer is not entirely accurate, since there is a lot of data that can match. Right next to the text, we have a option to give us a somewhat more detailed answer.
In Google sources are the basis of the search engine. Despite the fact that they have been giving more and more importance to their suggestions and less to pure links, Google continues to be a huge source of traffic for millions of web pages. With conversational search engines, the relevance of web pages remains a mere bibliography. The idea of the search engine itself is that you keep asking things and not that you click on those links.
Precisely one of the best of Perplexity AI is the ease of chaining questions, usually understanding the context of the information well. If we ask him about Apple’s mobile, he will give us an appropriate answer. And if we then simply say “and the laptop?”, he understands that we are referring to the Apple laptop.
As with ChatGPT, we can link multiple questions and the browser understands the context.
The rise of conversational search engines like Perplexity AI, which it’s still in beta, you can force a change in the way you search for information. On the rebound, the rise of these AIs can encourage websites to offer better structured information, so that it is easier for AIs to digest and that they link to those sites. It is something that already happens with Google itself.
Perplexity AI does not offer anywhere near the reliability that Google has, but as with ChatGPT, the proposal is attractive enough so that we do not see it as unreasonable to use these conversational search engines. Fortunately for users, this battle has only just begun.
In Xataka | Google has LaMDA to deal with ChatGPT, but it won’t take it out of the lab so as not to damage its reputation
