Perplexity additionally has a Deep Analysis instrument now, and it is powered by a model of DeepSeek R1.
In keeping with the announcement, the AI search engine’s new instrument does “in-depth analysis and evaluation in your behalf,” by crawling the online and compiles a complete report of its findings. If Deep Analysis sounds acquainted, that is as a result of Google and OpenAI each have their very own variations of the analysis instrument for Gemini and ChatGPT respectively — sure, they’re each additionally referred to as Deep Analysis. XAI’s new Grok 3 took some artistic liberties and referred to as its analysis instrument Deepsearch.
However whereas Google, OpenAI and xAI’s analysis instruments depend on their very own proprietary fashions, Perplexity makes use of a personalized model of the open-source DeepSeek R1. Perplexity CEO Aravind Srinivas confirmed this in a follow-up to a Feb. 3 submit on X saying, “can simply allow one thing like Deep Analysis at 10-100x decrease pricing, utilizing a customized model of R1.” A couple of week later Srinivas indicated that that is what Perplexity did by reposting replying, “carried out” with a checkmark emoji.
Mashable Mild Velocity
Perplexity is providing Deep Analysis freed from cost for its customers — with limitations. Non-paying Perplexity customers get free entry to “restricted variety of solutions per day,” and paying subscribers get limitless entry as a part of the $20 a month Professional plan. That is completely different from Google Gemini, ChatGPT, and Grok 3 which solely provide their analysis instruments to paying customers.
Whereas Perplexity primarily based its Deep Analysis off of DeepSeek’s R1 mannequin, as a result of R1 is open-source, it signifies that programmers can tweak and customise it for their very own functions. Final week, Perplexity launched its personal open-source model of R1, referred to as R1 1776, “that has been post-trained to supply uncensored, unbiased, and factual info.” It is a direct response to claims that the unique R1 censors responses criticizing the Chinese language authorities.
However Perplexity Deep Analysis isn’t with out flaws. Decoder found that the instrument inaccurately attributed the time period “stochastic parrots” to AI researcher Gary Marcus, when the truth is it was coined by Emily M. Bender in a analysis paper. Customers have additionally famous that Perplexity Deep Analysis provides outdated and inaccurate knowledge, which is an issue because the firm has touted it as instrument for funding and market evaluation. Srinivas mentioned in an X submit that they are addressing the problems, including “for finance particularly, knowledge accuracy is a should and excessive stakes.”
Nevertheless it simply goes to point out that hallucination for all LLMs is a persistent downside that may not ever go away, so use with warning.
Subjects
Synthetic Intelligence
DeepSeek
