Transcript for:
Изучение Perplexity AI и его вариантов использования

hi everyone thank you so much for joining I want to start it off by just telling a story about the founding of perplexity and so about three years ago our fanders had this belief that the foundational models are going to be increasingly competitive and it's going to become commoditized similar to how storage or compute is commoditized today and they thought the value would come from an incredible search uh as well as integrating with lots of different data and so what this was a contrarian Viewpoint um a few years ago and it was called an AI wrapper as you've probably heard that term and so what I'm really hoping to show you today is is how powerful an AI rapper um with an incredible search connected to your data can be so with that I'm going to jump in and share my screen so when we think about perplexity um what our primary use cases for the finance vertical is everyday search so our users who have been on here for over a year are really replacing Google with perplexity and so you see a question about news events happening and you want to go beyond the headline and so we're taking you from a world where you used to go to Google and search through 10 Blue Links and look at all those different sources click on them and read them um to a world where we go and take your query we break it up into about five different types of questions and then we go and and search the web in real time and find answers out there in the web and so here we've pulled about 60 sources and what we do is we're finding the best sources and ranking them all and then we're actually going in and feeding those Snippets from those sources uh that are ranked and we're feeding that to the foundational model that we believe will be best to answer the question and so we're really we're model agnostic and so we're going to find the best model at any given time and give you an answer so this question there's a new German Chancellor and so we want to know who is this person um you know what's their political background um which way do they lean looks like he leans right I'm a little bit more of a Centrist and what's their stance on the US I'm a US investor I want to know um how he is and so now you can see all these um articles are sourced in here so you can get a really good understanding as to like what the sources that are being used to generate this response and we've gone through 60 websites and what I love about perplexity is it's really it's picking the type of news that you would probably the Articles yourself um so you can ask this more questions you can ask follow ups about what would be the implications to US Stocks um and so there's a lot of information packed in here and so that's just scouring the web um and so our search is really powerful scouring the web um but I think what we're so excited about uh recently is that we've been able to do a lot more Integrations with data and so now I'm going to take you to the next step uh with perplexity is you can start to ask questions about let's say earnings you know there's a huge earnings call coming up tomorrow for NVIDIA and so I want to go through the Nvidia earnings um from the last period And I want to figure out what sort of they gave and here what I want to do is I want to connect to a third party data source so I'm I'm going to connect to fax set and so when we're doing this um this is a really powerful tool right we're not just searching the web we're actually searching an index of data that is a thirdparty integration like fax set and so we're going to search that we're prioritizing that data so we have like really trusted data and so you can see what it's doing is it's going to try and find the most recent earning call for um for NVIDIA and then it's going to go into that document open it up and read it and find out what sort of metrics they guided to so that you can be prepared tomorrow for one uh Nvidia reports and so to enable a product like this um you really uh just have to have a fact set subscription as well as a perplexity subscription there's no additional uh cost for it and so what we're we're really selling you here is is is you have an incredible time time save so you're getting uh like almost a cheat sheet uh for these earnings and so total revenue the number gross margins uh everything you need to do everything you need to have to be ready for the earnings call tomorrow so really uh powerful tool pause there if Jenny if you're seeing any questions in the chat and you want me to answer on this one yeah some folks are just asking um like is there a specif uh they just wanted see how it works with crunch base as well but I think you have yeah let's jump into crunch base too so so far um I've I've shown you just very simple uh Pro query and and so there's a lot of different ways to ask perplexity questions uh so for for most the time we like to use prir and that matches the right foundational model to the question you're asking but if you have like a more uh difficult question you want to ask more of like a re search style question and you also want to integrate with other data sources uh you can you can upgrade to uh deep research and so this is included in your Enterprise plan and in deep research you've probably heard about it um this is incredibly powerful tool I like to think about this as almost like a first year analyst um so incredibly smart and and what it does is it it actually reasons uh when it's thinking about your your question and I'm going to connect to Crunch base here because this is a question about um about recent fundings so I I want to figure out what's going on in the market right there's a ton of deals happening and I want uh perplexity to look at all these deals and figure out well what what is actually like the thesis that people are investing in it's a huge amount of work um and so deep research this takes about three minutes um so we're not going to sit here and and watch the whole thing I'll show you some other things as um as it's being done but the the important thing to understand with deep research is it's going out there with searching a bunch of sources um obviously those will include crunch base and then it's seeing the answer and evaluating that answer and then reasoning to think about well how can I improve that answer and then it's going to go do another similar search again and it's going to keep iterating and be really thoughtful almost like a first year analist would do if you gave them this project and so as that as that's running let's jump in and do some really fast searches right so Brer really quickly we wait for that to run a few folks were asking um how to access the fa set in the crunch Bas integration can you share a little bit more on um what access you'll need and then how to enable that yeah and so definitely reach out to us if you are a fact set customer already um what you can do is you can authenticate uh through perplexity and that will enable access to that you are subscribed to in fax set um so it's just a few clicks if you have that permissioning but definitely reach out um reach out to us and and we can we can help you on that path um so another like just just doing some quick searches here let's say you know the question we always get as as Financial uh experts is you know why is a why is x stock up today um and so you know AMT is I saw it uh in the news it's it's it's up to date and I would just want to figure out why so something you would have to go into Google again go through those 10 Blue Links figure out what's going on um with perplexity you get that information really fast so we're we're saving an incredible amount of time and that's the power of a pro search so you don't really need like a a deep research here um but what you're seeing is is they beat earnings uh they incredible Revenue growth uh cre organic growth even a better uh 2025 Outlook than the market expect um so that's powerful but if you want to dig deeper um lexity by being you know not so focused on the foundational models but focused on giving you maximum readability uh We've created profile Pages for for all these stocks um and so you can see over different periods you know you've got charting functionality in here really impressive but but what we really love is is we have live uh earnings transcripts flowing in here um and so if you were watching this call uh this morning you could watch the transcript flow through and You' be getting uh AI generated summaries at the same time so incredibly powerful Tool uh to be able to go through earnings imagine covering multiple companies at the same time and getting real-time earnings uh or if your client ask about a why is AMT up just being able to get the answer in seconds and so um I want to also show one thing that we recently launched that we're really excited about is we've created a dashboard uh for you kind of like a Google Finance um where you can come in and get caught up on everything that's going on and so what we've done here is is we've looked out at the market and seeing how everything's going seeing how the news uh headlines are and and we come up with a you know a sentiment analysis and then we have all the live earnings that are happening we have a written analysis that is updated in real time we have upcoming earnings top movers standout movers and explanations as to why these stocks are moving really powerful tool um which is SL finance and so now jumping back into that deep research query that we ran that usually takes a a few minutes to do um this is really powerful right so we're integrating crunch based results with a deep research uh response and so you know you're going to get your crunch based results here and you're going to see some of the larger series a deals um and you can click in and see more of this information here but like safe super intell IG raised a billion that's fascinating right um jumping into here now perplexity is going to summarize all this information that it's perceived um from unique data sources but this will also work uh just by looking at the web as well and it's going to come up with like what are the thesis Behind these um these all these fundraisings uh so you AI obviously a big theme U Frontier Technologies like Quantum Computing and Robotics huge theme and it's going to give you specific uh sighting back to like actual rounds in crunch base uh of companies that have raised sustainability and climate right um so many different interesting themes that are happening right now that are that are receiving a ton of capital a report that would have taken an analyst a couple days to do you're you're getting this in in three minutes um so that's like a huge part this about 80% of the use case for um for perplexity is this searching the web searching third party data but I want to tell you about um a new product uh that we've launched in the last few months that that I'm really excited about and that's called perplexity spaces and so imagine a space as a bit of a silo of the web and a place that you can come in and do repetitive tasks so let's say I'm building profiles a lot right um I want a space where I can enter any company and generate a profile in the exact way that I want it right I have a specific things I care about when I see a profile and so I don't want the standard web profile I want a profile that values my internal sources if I have them I want it to give an output that is very specific I care about the business description I care about the history I care about deals positive nudes and I'm really concerned about management um so I'm going to tell it this pre-prom and so this pre- prompt is great because it now is saved in this space that I've created so now every time I come back to this space I don't have to enter that pre-prom that I put in here I can just enter the company name think about if you're on your way to a meeting right and you really want to get a quick profile of a company I can do that here and so I think there was one that I saw um in that last page it was a safe super intelligence um let's just safe super intelligence startup um that one seemed seemed interesting I want to learn more about that they raised a billion dollars um so now when I hit this search uh it's going to take into account that pre-prom that I fed it and it's going to when it's going out and searching the web it's going to give me an answer um in that same format I requested um so here you know you see the things key manag M recent developments the things that I really care about um in addition to some more information obviously Source backed to all these different amazing articles and Brer really quickly in spaces can we clarify the types of things that we can Source um crunch Bas and fact set some folks are asking about U specific files and links as well yeah I think that's really interesting so like I just gave it a pre- prompt so now let's jump into some other ideas so you're like how do I do um links let's say for instance um so here's another space that is going to look at government websites because I'm sitting here and there's all these executive orders happening in the US at least and it's almost like too much for me to really process um but I don't really want like the bias of news um coming out and changing what's actually like coming out of the White House so what I'll do is I'll I'll ask perplexity what are all the sources where executive orders come out of and it'll give me these links um and so here I have you know white house.gov archives.gov so now these are like I'm siloing the web I'm saying there's like parts of the web that are actually more valuable uh in regards to this search than other parts of the web so now I can go in here and I can ask perplexity to um let's say go through the recent executive orders past this in here and tell me about the energy sector right energy sector analyst investor um and investment banker and I really want to know what's going on there and I I really want to Value these like truthful sites and a fun thing you can do here is you can say actually you know what the web I don't I don't know if I trust that for this information I only want the links so when you select these links I'm only looking at the links from this space and I actually care about information that's come out in the last month so another setting you can do is set the the timing of these articles so because itly have to make a decision as to what is recent and I'm going to try another mode here which is uh deep seek reasoning um and why I chose deep SE reasoning um is is perplexity is going to uh give you the best models at any given time so right now we're seeing like deep seek is incredible so we're going to give you the option to actually like directly uh grab that you know in two weeks um when grock 3 releases the API that's probably going to be the best model and so you're going to be able to choose that as well so by being model agnostic we're we're able to always give you the best model at any given time and you don't have a lock in to one particular model because it's so competitive and the costs are getting so cheap you really want to experience every single one uh deep seek is is is is a incredible one right now um because it does a lot of reasoning it actually like thinks about your query and and is making bets on which directions it should take in answering your query um so it's grabbing all the sources yeah just want to add that point I know a lot of folks ask um we don't train on your data especially since we're not a model company and we really just want to make sure that we're providing the best and most accurate answer for your questions um I know a lot of folks have uh questions about like do we integrate with cap IQ pitchbook Etc and we're only seeing crunch Bas and fact set right now can you share a little bit more about um our specific Integrations Brooker yes so we've launched uh fax head and crun space uh in the last month or so and we're seeing amazing feedback from that we're seeing a ton of interest in it and and I think back to my original point is is we're really focused on search and integrating with data so i' I'd love to hear like the top ones that you'd want you can just throw it in the chat what are your favorite ones if it's pitchbook or Bloomberg and and we're working we'll work to build towards that we're really we're really responsive to to what you guys actually care about and and we're happy to build that but those are the ones we have now and then you'll see we'll be launching some more um pretty soon um so this one what I love about this is you know we set the index ourselves uh so imagine you could do this for let's say if you're looking at um an area that's highly regulated like the payments industry and you're a payments consultant uh and you only want to get information from the payments industry well just set your index to the pay industry um and then we'll we'll be able to prioritize those sources or only use those sources and so you can see it found an incredible amount of information 29 sources from these government we websites uh and you can see all these things are are fairly recent uh so there's energy dominance cancel fossil fuel development grid reliability regulatory roll backs like there's a lot of really interesting things here you can dive deeper ask a followup um but it's I think this is one of our Our Favorite Things um here I think another one um that we can go into in the spaces um just coming back to that idea of that pre- prompt um interested in is you know you can create let's say if you're a banker and you have this repetitive task that you're doing constantly uh let's say you're looking at um deals and you got to put together Tombstone on this deal and you can do this for anything you can get like incredibly detailed in your pre- prompt and have like very very specific outputs you could even upload a file that has this and ask perplexity to always use this as a template and so now when you search something uh like J&J's acquisition of int cellier it can go and use fax set it can go and use the open web but you're getting the exact information that you want and so you can have this um for for your client for your your boss and at the end you could export this to a PDF you could just copy the text down here um and then put that wherever you want so you can really play with how you want the formatting to be within perplexity you can ask it to generate tables all in that pre- prompt so repetitive tasks are saved here um and so these yeah one more thing on sharing broker you can share this with your teammates right you can choose to share it either public or even within your own team so now um what I think is really compelling is like you created this space this profile Builder and so now you can share this with your team because now you created IP and so now they can see that uh and so now you can build these like all these templates and and start to build IP in your organization and start to attach more documents to it and then you start to share that knowledge with your organization so you could invite collaborators up here in the top right like I just did for Jenny uh you can Al and so when you invite a collaborator now they're capable of sharing or and editing your prompts and then there's a difference between collaborators and viewers um which you can then share up here in the top right so if I want to share this um with my organization I can do that and now Jenny is able to see it uh so but she's not able to edit so very different and and just to be clear whenever you write um a prompt like this our default is that it's private uh and so we we're just given the industry we're in um we we value privacy and so we're really like everything is going to be private so you just got to make sure if if you do want to share everything in here you'll have to share these individually um now jumping in so we've gone through a bunch of levels right we've gone through the basic functionality around searching and then gone into spaces where we can do pre- prompts and then we can add links and so tying it all together uh is this space uh let's call it an investment analyst and and so this is like the next layer is well I want to add a bunch of internal documents and so you can manually upload documents and we also have the functionality to connect to uh Google Drive and and SharePoint which will be launching this week which is fascinating so you don't have to manually upload these um but what you can do is you can add a bunch of your investment memos um and so here I've downloaded arcs investment memos because they're public and then I've added it to this space and so I've got all these investment memos in here and now I'm able to uh index these investment memos and so I can ask questions like go through all of arc's memos on the web and in the space and then build me a new memo for Uber based on all the public available information and write it in the same style I can even tell it to make a decision as to if it would invest in U so you're getting like a view as to arc's Memos before they even publish the memo based on their historical memos formatted in the exact same way um as their as their prior memos and so you can see their thesis you know they love self-driving case for Tesla and so you can see that they're actually coming up with a bu rating on this based on a similar um similar thesis the jeici cded network margin expansion growth in other markets amazing momentum so huge cash flows it's just like a really compelling thesis again back to that point is we're almost at this like first year analyst um first year analyst uh reading these reports and all connected to live pricing coming up with valuation um so this one's kind of combining all those layers of web third party um and your inter to build something really powerful again you can export this to a PDF um or or copy it here so pause there on the internal data see if there's questions yeah I I see a lot of folks do have to run um this will be recorded we we'll also share these spaces in a follow-up email um but really quickly once we share those files for sources Brooker a lot of folks were asking what is the security and privacy on these and how do we make sure that this isn't necessarily shared um and we keep every our customers files private yeah that's a great question um you know that's like a I think the core offering for our Enterprise clients is is really privacy and security and so we've built this product um on that and so just going to take you into the controls you have and so obviously when you when you're adding um files you know we're storing it encrypted and stored um that are Keys unique to you um and but there are some differences within the product and how we store it uh when you when you add documents um here let's say from a local file um this is in your search we retain it for 7 days now when you add documents into a space um we retain this uh the documents like in here we retain this until you tell us you want to delete it um and some organizations um you know may not be uh that comfortable with it and so what you can actually do is is within your organization there are um permission settings that you can change um let's see we won't go through that right now but there's permission settings that you can change where you can actually like permission um certain users to be able to upload documents you can um permission users to be able to share threads externally um you can allow users to save any of the queries they've had so like data retention you could turn off any data retention so it just like Google it's not saving anything it's deleting it immediately so there's a huge amount of customization you can do as well if your organization wants that great and also for folks who do have more questions about how we make sure that your files and your searches are secure um we have everything shared on trust. perplexity doai I'll also share that in the followup but um we definitely have a very comprehensive overview of how we make sure that everything is secure and so just maybe is Jenny compil some questions I just I'd love to talk to you about some of like the use cases we see that really resonate um with Finance customers U and so my background I was an investment Analyst at kotu and Citadel and I love to think about IPOs coming out um and and when I was chatting with a client uh they shared this prompt with me that I thought was incredible I'm just going to type it in here um I'm going to use um R1 here and I'm I'm going to use the the web to be able to figure this out so just as this runs this is is really like where you should be thinking about um how to get the most out of perplexity now is like what would you send in an email to an analyst to be able to do this work um and and really spell it out and the more detail you can provide like first make an assessment of who the peers are I'm trying to build this valuation model fororo then find the valuation metrics then find what the revenue growth would be then figure out the proper valuation metrics to use for my company based on the metrics you found um and then make an assessment of it so like a huge amount of decisions that need to be made here and I'm giving it a little bit of Clues right I'm not leaving it totally in the dark I'm saying Robin Hood should be appear um so now we're we're reasoning and so there's um 86 sources in incredible amount of knowledge it's found out there on the web um and it's starting to make like a ton of decisions uh to be able to do this right it has to pick peers and it has to figure out the valuation of the peers like there's so many steps involved and you can kind of watch this flow and as you're watching this flow that can help you define if you want to have like a different prompt because you can see where it will stumble um so if you think that it's it's starting to choose the wrong Pro wrong peers then adjust that right chose Charles Schwab as a really close Pier um it's also has other ones that are private like public and Weeble but you can see it's like it's taking these steps figuring out all these metrics things that would just take you so long to figure out the revenue growth what matters I didn't really tell it exactly which metrics it should use um and then it's doing math in here right it's it's it's coding on the Fly um and it's figuring out well what's what is the the right valuation um and so it's giving you like a four to 5x EV to um to revenue 5 billion IPO and so now you can look back and say well does this sound right uh what would I do differently and then you can kind of be more of an an objective Observer uh as to like how it's coming up with it how did this first year us think pause for some questions I see a lot of love and excitement for deep research um I know some folks were really interested on how sources are determined like how do we know that these are the best sources for my question and I think it would be great to show like hey I don't really agree with certain sources let's remove it and potentially regenerate the question yeah that's a great one so a lot of sources in here I think the first thing to notice is you can go through here and you can kind of see like which ones it's using the most like it's getting data from uhoro Direct that's a pretty good source uh I'm getting data from Robin Hood Direct that's a pretty good source maybe but this like Wall Street season um and the stock broker one I don't love and you can see so like while we may have found 86 sources we feed all those 86 sources to the foundational model that will best answer this question and then that foundational model is given um a snippet of text from The Source along with our scoring of the source and we base it on like 50 metrics to evaluate the source and so it's fed both those things and then the foundational model is making a decision as to whether it include that source so while it found 86 the foundational model may not actually like believe all 86 and so it's only going to use um a dozen and you can see which ones are here but if you wanted to remove these and then you can rerun it um if you don't trust those sources uh you can you can do here and then you can also control the sources as well we also got some questions on prompt engineer and also considering we have a really International crowd today um can you share a little bit more about kind of making sure div we're diving into other markets and do we clarify that in our prompt or is that something we could do with links and sources y um so I would encourage you yeah so there's a bunch there prompt engineering I think like what we're trying to do is we're trying to do that work for you we're both trying to do the work of the prompt engineering so we're taking this query and we're building a research plan to come up with the best searches that will get you the right information for this so we really want you to just think um and interact with what feels the most natural and we'll do the work of prompt engineering and we'll do the work of finding the best model um so I can give you some tips as as to ways to do it um just after um as to like what are like good questions to ask but I would really think about uh talking and emailing a first year analyst when you're trying to come up with the right prompts um and then yeah International companies always tricky right the SEC data is super organized uh and and really easy to get feeds to international data like we were chatting with a large Bank in the UK and they are trying to get financials data and that's not as easily found um and so what is best you could create a space on a specific set of companies and you can go through uh let's see like the government action one you can go through here and start grabbing uh specific like IR websites let's say you have 10 companies you want to look at put a bunch of those uh IR websites here and so we'll be able to prioritize that so you get uh truthful financials so I definitely would play with the links and then um we're also definitely um thinking about how we get that international data as well that is in in our product road map because most of our us users are at outside the US um I know some folks were asking can we Blacklist certain sources I we can't necessarily Blacklist but a space is definitely the best way to focus in your search if you want to allow specific links yeah um we're also getting a lot of questions on how do we pick the best model like why why do we allow our users to choose different models in Pro and Enterprise Pro yes that it's a great question question um the way I think about it is like 90% of the time I'm in Pro um that's really crazy fast we pick the foundational model that best suits your answer um that's amazing what you'll see here is um I like to think about is like it's the hot new models um that some of our users uh really want to test out and use and get a feel for um so like you know deep seek and 03 mini these these models are being discussed a lot like when grock 3 comes out with the API in a couple weeks we're going to add that here and then deep research is is also uh something that's been um been really powerful and people want to test out so I would say think about um you know probably more like how deep you want to get like if you're asking what the score of the Super Bowl was an auto is like super fast um like that's basically if you're just where is this restaurant what's on the menu at this restaurant um then deeper is like Pro search takes a little bit longer and then even deeper than that is is deep seek and then even deeper than that is is deep research um and then so we we kind of give you that option here and if you really want let's say you're like convinced grock is the best um you can actually go into your settings and set the model um over here instead of it being Auto uh you can set it to like grock or Cloud 3.7 nice I also know a lot of folks are asking can um perplexity search things that are behind a pay wall um technically no that that that would be hard unless you're authorized into through like a fact set or a crunch base we do have some like publisher Partnerships uh with companies like LA Times or Dar Spiegel um or time where you could get access to different content um and we're trying to build that but it's as the product stands today it would be hard we're we're working on some Creative Solutions like we launched our own browser yesterday um to be able to tackle that problem great um yes the recording will be shared I know folks uh are asking we will also share these prompts and spaces and I know I think a few folks ask um have been asking on 10K sentiment analysis I believe that you've built a space before and added them as file sources broker uh yes you can there's a lot of ways um just showing a fun prompt in terms of prompt engineering um but there's a lot of ways to do that in spaces I would um upload like we created a fun space um for Warren Buffett recently um that does that sentiment analysis so um this one so what you can do is you can go and upload a bunch of 10ks um in here and you can start to ask questions um about the the document so you could upload Warren Buffett's most recent um shareholder letter and and ask it to go through so I could let's see if this will work if I connect to web I'm this and so here I'm looking at the web and the documents in the space I'm going to do a reasoning one here because it's a little bit more complicated um and so let's see this does but yeah shareholder sentiment analysis is is a favorite of ours for sure and then as this is running we're just going to flip fall because I know we like two minutes left and I just want to give you some ideas for for fun prompts if you're learning about a new industry think about like SWAT analysis pteri forces uh things that can cue perplexity to give you a better answer um think about using tables right like go through you want to go through the fomc minutes and compare to Prior prior minutes um my summary use a table we catch typos Brooker is just being diligent with keyboard yeah it'll it'll figure out the typos and clean it up for you so this one's running and so you can kind of see the change in in um in sentiment over it and the things he cares about uh running in one minute left you can see how I gave it this uh prompt um so did an okay job I think it's missing one meeting um but you can see it's filling it's deciding what are the key metrics it's is crazy decision and sourcing each of those just a huge amount of time any other questions um I know someone asked how many files you can upload you can upload up to 500 per space and one uh section of questions we didn't quite get to is how can you get to that Finance dashboard and a lot of folks are asking if you can upload your own portfolio to analyze that yes so just type in finance click here here's on the finance dashboard uh we're going to add personalization it's a little clunky right now you go in here p and you can add your watch list here so your stocks uh in crypto you can add them here and then um in about a couple weeks you'll see that show up in the dashboard