roadrash99
Level F
What do you guys use this service for, if you don't mind me asking. I'm interested in the real world use cases for AI and would like to learn more about how you people are using it.
Perplexity AI Pro
- Thread starter roadrash99
- Start date
mutant
Level G
Similar, but not at the level of chatGPT and Gemini 4 etc. But its decent for personal usage.
TheBark
Disciple
It’s apparently better than GPT4 in some benchmarks.
Pat
Level F
Although Perplexity has it's own LLM, the Perplexity Pro subscription provides access to GPT-4, Claude 3 etc. which are regarded as the best in the industry. So, technically speaking it is the same as "ChatGPT" (I would say slightly better since it offers different LLM choices)!
Already using it for work. Used it extensively over the last 2 days for research for two client decks. Good enough information with infographics and citations, saved me quite a few hours. If it maintains this level of quality or gets better, $200 a year is nothing, heck even $500 dollars is nothing since it’ll save me countless hours in research and intel gathering.
Last edited:
Some examples of how I’ve used it the last 2 days - Market intelligence on what peer group companies or competitors are doing, research on potential locations for cheap talent, near-shoring location shortlisting, how much are CXOs being paid and what kinds of compensation instruments are being used in competition , analysis on a particular industry, economy, talent skill set. Looking up infographics, creating financial data tables of industry and financial performance of multiple firms.
I’m more and more using this instead of Google
chyawanprash
Level E
Looks like I need to upgrade my brain to be able to extract something meaningful out of Perplexity. So far it has given me similar results as Google.
I tried asking it for customized gym equipment. It couldn't explain beyond basic blog level stuff. Instructables has tons of details on such stuff but it gave references to irrelevant pages. Images and video results were also all over the place. I then gave it a couple of reference images and it gave bullet points like a low-level blog but then again, those points were vague. Playground generated illustrations showed that it doesn't understand physics and the human world. I was very amused to see that.
While using the free version of ChatGPT, it felt like I needed to possess 75% of the knowledge beforehand to be able to reach 90% with its help. It was consistently spewing wrong information related to acids. All that info was pretty basic and I was able to catch it. It was also wrong about basic queries related to Windows, Excel, batch scripting etc.
I should be able to get good results from the paid version of Perplexity, as now I have access to GPT-4, Claude 3 etc. I want to use it for stuff that I have no prior knowledge of.
Right after buying the subscription from TE, I tried asking it about some TE feedback related stuff. It gave generic response that'd be applicable to any forum. Relevant information is already available on couple of threads on TE and I thought it'd be able to pick that up and summarize it for me. But it couldn't. I then informed it about what it was missing in one line. Like an obedient kid, it wrote that it'd give out the correct procedure whenever someone is going to ask the same question again. I was surprised to see that it laid out all the steps with crisp explanation better than what TE already has.
Anyways, I tried asking it about the same issue after a couple of hours from a different PC and login. This time around it gave me steps to use javascript to do some mumbo-jumbo.
How reliable is it though at the end of the day? All the LLM models hallucinate to the extent of even making up citations, leading to some lawyers being persecuted for quoting cases that never existed. Checking every source for accuracy is also quite the job.
I remember once Copilot stating something that I knew was factually incorrect and the citation didn't even include the information that was quoted in the reply.
Google at least scrapes the actual content so you know it exists at source, the LLM just makes it up based on probability.
That’s par for the course. I have one of my reportees go over the narrative and citations, if it makes sense. but its great for story boarding a deck and figuring out the key messaging.