Transcript: I didn't realize OpenAI had new embedding models. It's probably useful to re-embed with these models, and then also... re-embed with the models, and maybe even trying other models that are not specifically OpenAI-based.
82.86% similar
Some people are realizing that embeddings alone are insufficient, and they are proposing different approaches to address this issue. A few suggestions include converting queries into keywords and conducting searches based on them, utilizing indexed full-text search databases, incorporating knowledge graphs, and implementing named entity recognition and traditional pair extraction techniques. A hybrid approach that combines all of these methods may be necessary to effectively tackle the limitations of using embeddings alone.
82.58% similar
The ability to create embeddings at different scales would be very beneficial, particularly for finding relevant sentences within a document. This kind of processing could be even more useful than the initial summarization. It might involve a manual process, but it's a direction worth exploring within the context of search capabilities.
80.98% similar
The article "The Bitter Lesson" shared by Raphael emphasizes the idea of relying on computation to achieve greater capabilities in artificial intelligence, rather than complex feature extraction methods. It underlines the notion that the accelerating pace of computation enables more significant advancements in AI. The possibility of tackling problems by increasing computational resources is highlighted, particularly in the context of contextual AI. The article suggests that models like Mamba, which employ state space techniques, may offer potential avenues for this approach.
80.50% similar
The speaker is considering moving Whisper to their computer to save costs, as the audio component is expensive. They mention that the cost is mainly attributed to the audio, while embedding is negligible. They express surprise at the high cost of audio, which seems unjustified given the amount of audio content. The speaker is contemplating how they could potentially reduce the cost by moving the audio component to their computer.
80.46% similar
The speaker is contemplating creating a micro social network, motivated by both marketing and curiosity, to explore what a small-scale community might look like, with the idea of bringing together a select group of 25 to 100 people to experiment with the concept. They are considering sending out a survey to gauge interest and see if people are willing to collaborate on developing the platform, suggesting that even with basic functionality, like comparing user embeddings, the project could rapidly evolve. The speaker hints at a profound potential in this endeavor and expresses a keen interest in delving into it, as well as providing the means for participants to create their own websites and forge connections. Acknowledging data as a major challenge, they note that providing tools for the community could bring the project closer to addressing this issue.