Announcing Perplexity Ask, a new search interface that uses OpenAI GPT 3.5 and Microsoft Bing to directly answer any question you ask. perplexity.ai discord.com/invite/kWJZsxP…
Inspired by OpenAI WebGPT, instead of displaying a list of links, we summarize the search results and include citations so that you can easily verify the accuracy of the information provided.
@perplexity_ai The goalposts of AI have shifting again. "Find out what there is to know about X and cite relevant sources" would have been my typical Turing test question. Perplexity AI does this effortlessly.
@perplexity_ai 🤣🤣🤣 AI not smart enough to catch that FTX’s troubles started when SBF and Alameda stole their customer deposits? (Long before withdrawals were made by Binance). Will fix it. What happened with FTX? Answer: SBF stole billions.
@perplexity_ai Love the links to real sources!
@perplexity_ai You should monetize with USDC micropayments using @handcashapp. Penny per search!
@perplexity_ai Great work! We've also released an answer engine that combines GPT-3 with human-validated facts + Google search. We are sourcing and fine-tuning our system for essential local information. You see it here: maryanne.faqx.com
@perplexity_ai Love seeing how things are going with natural language question answering. The hallucination / agreeability we're seeing these days is a lot: > what are the opening hours for the moon rock giveaway at the Museum of Science in Boston?
@perplexity_ai Imagine what @perplexity_ai could do when it changes Bing for a real search engine
@perplexity_ai this is such a nice post! goes to show how the first version can be evolved to something better and greater
@perplexity_ai @goodside How much are your OpenAI bills?
@perplexity_ai Great idea and execution. One challenge with combining results from different pages is disambiguation for different people with the same name. E.g. below with "who is Laurent Bonnet?". You get a superman who did all those things, when it's in fact the bios of 3 different people.
@perplexity_ai Interesting. After a few tests, it is difficult to trust the info delivered. Sometimes the system only mixes partial information to continue the narrative. Sometimes it takes a partially matching source, assigning facts to the wrong entities.
@perplexity_ai Very cool. I think this would be helpful to develop web scraped datasets (e.g. an up to date list of where doctors treat patients by hospital / clinic has been hard to create in the past.) Are you planning on releasing an API?
@perplexity_ai ok so can we talk abt the ui/ux for this ☕️
