AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
3min snip
20VC: Perplexity's Aravind Srinivas on Will Foundation Models Commoditise, Diminishing Returns in Model Performance, OpenAI vs Anthropic: Who Wins & Why the Next Breakthrough in Model Performance will be in Reasoning
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
Exploring the Enigmatic Nature of Model Capabilities and Reasoning Advancements
The quality of a model's performance is heavily reliant on data curation. The belief in the verticalization of models for specific tasks is challenged by examples like Bloomberg GPT being outperformed by a GPT-4 in finance benchmarks, indicating that the magic of these models lies in their general-purpose merging capabilities. These models are not trained for specific tasks but rather on a diverse set of data that enables them to abstractly pattern-match and excel across various domains simultaneously. The remarkable reasoning abilities of these models remain enigmatic, as it is unclear whether training them on specific domains like math or code would enhance their reasoning skills. The effectiveness of these models in practical production scenarios stems from their innate ability to catch patterns on the fly and perform tasks without explicit training for those specific tasks.
00:00
Transcript
Episode notes
Get the Snipd
podcast app
Unlock the knowledge in podcasts with the podcast player of the future.
AI-powered
Discover
highlights
Listen to the best highlights from the podcasts you love and dive into the full episode
Save any
moment
Hear something you like? Tap your headphones to save it with AI-generated key takeaways
Share
& Export
Send highlights to Twitter, WhatsApp or export them to Notion, Readwise & more
AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Discover
highlights
Listen to the best highlights from the podcasts you love and dive into the full episode