Search needs to help people complete the task, not just find answers

Transcript

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

1. 27:24

   Okay, let's talk a little bit about the future. I've seen your most recent launches are kind of like in different directions, more verticalized or more specific around shopping or other things. Where do you want to take it? Like today, my view of perplexity is a more intelligent Google search that's really useful in certain scenarios. What do you want me to think of it in three or four years? If you go and research what's the best sweater to buy or which is the best hotel to stay in this location, perplexity will give you a great answer.

2. 27:58

   But where do you actually go? You go and fulfill the demand. You go to Google. And who gets credit for that monetarily? Google us. Google. We got nothing. Maybe we get you a pro subscription, but then someone else will undercut us and give it away for free with cheaper models or whatever. They have bigger cash results. So the challenge is you want to be one place where people can have the end-to-end experience.

3. 28:22

   they start with a problem in their mind and they seek your help you give them the answers and you also help them fulfill the action it's difficult because uh people think like at the app if you have an answer of like oh like what watch does uh bezos wear i think he wears some omega something um i personally thought it would be amazing if it not only give the answer but it also had a product card for the specific omega watch with a buy button and i just click buy and it's done but there are other people in the world who think that's an ad it's not even an ad right they think like that company is paying us to do this so this is where like some of the tension with uh the early adopters who love the ad free informational experience with like what is actually needed to get mass market and really be useful on a daily basis comes from there are so many

4. 29:18

   other things like checking the score of a game or quickly getting to a website uh if you just wanted to get a docs link of an api or if you just wanted to go and book a flight on united the answer could just be a link the answer could be a the weather for tomorrow that's the temperature or someone's age should be like you're going to type like elon musk network you'll just get the answer in like less than a second on google right and perplexity it'll pull the right source maybe it might be more accurate than google but people don't care about like some of those minor details so what you need to build is this amazing orchestration of small models uh typical knowledge graphs uh widgets lm streaming answers and more complicated multi-step reasoning answers but user doesn't care like user's not going to tell you like when to use what you decide that ai nobody talks about

5. 30:12

   like when to use WOD, that sort of router, that orchestrator. I think that's the hardest thing to build. And whoever builds that and can operate that at a scale of a billion users and also knows how to monetize some of those queries really well is going to be the next Google. Because they'll have the search bar. They know exactly what to do. They'll go and ask clarifying questions. It truly understands the user and also does tasks for you.

6. 30:38

   and also lets you like surf the web in a typical way, all in one experience. You could even argue maybe nobody will ever be able to build this because it feels like a daunting task. But I could say whatever Google has already built is the closest system to something like this. Agreed. So the next generation of this clearly can be built. You just have to like persevere