Perplexity CEO - Any questions?
Added 2024-02-08 11:56:25 +0000 UTCAs always, first dibs on questions for my interview guests goes to you guys. And I am lucky enough to able to have Aravind Srinivas, Perplexity founder and CEO, formerly of OpenAI, as a guest later this month. No guarantees for any question but the most upvoted ones will get a very strong look-in!
For those who don't know, Perplexity are being talked of as the Google-search-replacers, though of course time will tell:
New York Times- Can This A.I.-Powered Search Engine Replace Google? It Has for Me. https://www.nytimes.com/2024/02/01/technology/perplexity-search-ai-google.html
Fortune: Jeff Bezos–backed AI search startup’s CEO says ‘Google is going to be viewed as something that's legacy and old': https://fortune.com/2024/01/06/jeff-bezos-nvidia-funding-round-ai-search-startup-google-rival-perplexity/
CNBC: Is this the end of Google Search? How the giant could lose its lead: https://www.youtube.com/watch?v=luWf6Zfyvms
Comments
Long time user of Perplexity Pro. The question is do they plan to use a better (like OpenAI's Whisper) speech-to-text model instead of Siri on iOS for example for voice input?
Shikhir
2024-02-19 16:46:22 +0000 UTCI really like Perplexity for research, and it works great in English, but when I have questions in German, or about Germany-specific laws it tends to translate them into English and thus comes up with results for the English laws instead. Are there plans for multi-language or localized searches anytime soon? (I have not yet subscribed to Perplexity because of this, so I don't know if the Pro Version doesn't have this issue).
spaceemotion
2024-02-16 22:01:27 +0000 UTCWhat books helped shaped how you build Perplexity or useful for 1st time founders?
Jaison S George
2024-02-14 21:41:09 +0000 UTCOh and all those requests cost $0.25. Total. Imagine if some of that advice really helps a student? That’s amazing return on investment. Not a question just another testimonial. “Thanks for keeping API costs low.”
John Weisenfeld
2024-02-09 16:21:42 +0000 UTCOne big reason I use Perplexity above anything else is the footnotes. Now I realize they are a leftover from Google Search but the links really feel like Perplexity have solved the where-did-AI-get-its-answer problem in other chat bots. But that’s the question has Perplexity made progress on the source attribution problem. Do they feel that provides competitive advantage. It might in education markets…
John Weisenfeld
2024-02-09 16:03:01 +0000 UTCLast weekend I tried all the APIs (ChatGPT, Anthropic, Bard—well last weekend was internet ages ago :) ). The only API which I could get to work was Perplexity and I used it to generate 1000+ custom career advice emails to students at my high school. Not a question just a testimonial. Thanks PPLX folks. I wrote this up on my LinkedIn. (Http://tiny.cc/lynkedyn
John Weisenfeld
2024-02-09 15:59:27 +0000 UTCThe pplx online models are very useful even on the API. why do you think the industry hasn't been doing much in this regard? Do you folks plan to invest more on that? Currently, It's not great with Chat on the API.
Tarun Pothulapati
2024-02-09 11:38:09 +0000 UTCWhy do you think Gemini Ultra failed (did not even reach the level of GPT-4 in many aspects)?
Ruslan Lagashkin
2024-02-09 07:07:41 +0000 UTCUpvoted👍
Brian Crabtree
2024-02-09 01:42:44 +0000 UTCThey seem to be on a roll regarding hardware partnerships, is there a lot of room for AI powered hardware beyond smartphones?
Anouar Mansour
2024-02-08 22:07:08 +0000 UTCIndeed, if Perplexity started getting successful it surely wouldn't take long for something like Gemini to copy it and then win with an incumbent advantage.
Poss
2024-02-08 21:51:37 +0000 UTCHow do they determine what is true? Replacing Google is not just a market disruption, its also societally usurping the source of truth for most people. This comes with a huge responsibility, since any manipulation of answers is effectively manipulation of society. Google uses an ontological knowledge graph to keep track of what is true. It was originally developed from Danny Hillis's Freebase project which tried to map all facts in reality. Danny now believes that approach is doomed, and his new knowledge graph just records claims about reality, rather than statements of fact. This contextual truth is also adopted by search companies like Google to an extent, since their products change answers depending on the country you're in. i.e map borders reflect the opinion of the searcher's government, not international law. Will Perplexity also enforce the incorrect and harmful opinions of dictators on their citizens? Or will their AI have more objective and consistent ethics to ensure alignment? If they are successful, the answer to this question may determine the future of humanity. Various world leaders have already claimed US companies use algorithmic social manipulation power to undermine governments in the Arab Spring. A hyper-convincing personalised AI will surely be capable of overthrowing governments, or of getting banned for trying.
Poss
2024-02-08 21:43:08 +0000 UTCI was out driving yesterday, and got caught in literally insane traffic. It turns out, Biden was in town in NYC, and I didn’t know because I work from home. At the time, I didn’t understand why, so I figured I would try Perplexity, and it could not explain to me that Biden was in town… I was stuck there not actually having any better answers. Isn’t Perplexity designed to provide insight during these types of experiences?
Stefan
2024-02-08 18:08:19 +0000 UTCAt the moment Perplexity is refusing to do anything online only give information. Will this change with the partnership with Rabbit and you also will be kind of assistant or the use cases will always be diferent?
Arek Stryjski
2024-02-08 16:39:26 +0000 UTCWhat the plan is for the future when they replace Google? The free search will not be available if adds model colapse, the information are much more scattered in the internet without adds or behind the paywall. How indexing of Internet or acquisition of informations will be payed off?
Arek Stryjski
2024-02-08 16:35:50 +0000 UTCExectly, for this type of question currently the answer is: "Oh no, I would need to look at this and then something else, and maybe something more... Do it yourself man!" :-)
Arek Stryjski
2024-02-08 16:29:30 +0000 UTC@GilSyswerda hey Gil, there was an earlier interview with Aravind where he mentioned that their plan on making future profit will be from a Workflow store where people will sell their training workflows they create and Rabbit will get a cut. Which I think is pretty cool...would be great to hear more info on the Workflow store 👍
Brodie50
2024-02-08 16:01:25 +0000 UTCWould love to hear an update on shipping dates for the batches. Also, would be great to know if there is any way to set other methods for the Talk button? For example if I am sitting at my PC and want to ask it something, is there a way to setup a keyboard key as the Talk button?
Brodie50
2024-02-08 15:57:23 +0000 UTCI love Perplexity today. It has a unique offering and it's arguably the best "Answer Engine" (as opposed to Search Engine). But it seems like what makes it unique today could be incorporated into future versions of the big foundation models. Today, the landscape is a bit fragmented. Do they see it continuing this way or will there be consolidation where, as Sam Altman has said, instead of many models to do many tasks, there will be one model to do all tasks? And how is Perplexity planning to evolve to fit whatever scenario they see unfolding?
Steve DeMoss
2024-02-08 13:28:18 +0000 UTCAs personal AI agents become a thing, why wouldn't users simply ask their AI agents for information?
Gil Syswerda
2024-02-08 13:27:16 +0000 UTCWill they bias the responses to the beliefs and biases of their users, telling them what they want to hear? If so, how will they learn about their users?
Gil Syswerda
2024-02-08 13:25:44 +0000 UTCIn 2-3 years, what will their main revenue sources be? Subscriptions? Ads?
Gil Syswerda
2024-02-08 13:23:42 +0000 UTCHow much effort are they putting into things line Ui/UX and what's their attitude about this, given that what really made GPT-3 explode was the ChatGPT product itself instead of the model
Eli T. Drumm
2024-02-08 12:35:19 +0000 UTCI would love to hear how perplexity is internally evaluating the outputthat they are generating.
SteveHaupt
2024-02-08 12:09:39 +0000 UTCTo what degree do they see website crawling as being a vital part of serving information to users int he future? Will search engines have to crawl every website on the internet to rank high quality content for the vast majority of queries? Or will they be able to crawl say the top 1000 (10000?) websites and rely on LLMs to respond to anything outside the scope of these sites with generated answers? Obviously these would be answers without citations. It's more of a question as to whether a model like this would be a viable way of competing with a Google and its ridiculous infrastructure. Or do they do this already?
Cooper Patton
2024-02-08 12:09:14 +0000 UTCExcellent that you got Aravind. My question is when will perplexity be able to synthesise data in a more advanced way. For example the data exists to answer the following question: "When did Lewis Hamilton and Arsenal last both win on the exact same day" that is something it can't answer yet.
Rakesh Murria
2024-02-08 11:59:50 +0000 UTC
