In this episode of Gradient Dissent, Denis Yarats, CTO of Perplexity, joins host Lukas Biewald to discuss the innovative use of AI in creating high-quality, fast search engine answers.
🎙 *Listen on Apple Podcasts*: wandb.me/apple-podcasts
🎙 *Listen on Spotify*: wandb.me/spotify
Discover how Perplexity combines advancements in search engines and LLMs to deliver precise answers. Yarats shares insights on the technical challenges, the importance of speed, and the future of AI in search.
✅ Subscribe to Weights & Biases → bit.ly/45BCkYz
⏳Timestamps:
0:00 - Introduction
0:55 - Denis describes Perplexity as an answer engine.
2:24 - Discussion on using third-party APIs and in-house infrastructure.
10:32 - Choosing between In-house vs. outsourced models
16:48 - Evaluating the quality of results and using LLMs.
18:39 - Building a classical search engine and custom parsers.
20:53 - Latency and quality trade-offs in providing answers.
26:21 - Handling controversial domains and providing unbiased answers.
30:07 - Importance of hiring diverse annotators and their influence.
33:20 - Denis's story about pitching Yann LeCun.
38:24 - Early signs of success for Perplexity's gen AI application.
41:33 - The hardest parts of scaling up the application and maintaining focus on speed and accuracy
43:00- Closing thoughts
🎙 Get our podcasts on these platforms:
Apple Podcasts: wandb.me/apple-podcasts
Spotify: wandb.me/spotify
Google: wandb.me/gd_google
RU-vid: wandb.me/youtube
Connect with Denis Yarats:
/ denisyarats
x.com/denisyarats
Follow Weights & Biases:
/ weights_biases
/ wandb
Join the Weights & Biases Discord Server:
/ discord
2 июл 2024
