))
Overcoming Search Challenges in E-Commerce with Alex Gunnarson and Zach Hudson of Onton
Manage episode 443088936 series 2903792
Welcome to AI Innovators podcast with your host, Rob May.
In this episode of the **AI Innovators podcast**, I chat with Alex Gunnarson and Zach Hudson, co-founders of Onton (formerly known as Deft).
Onton was born out of Alex's frustration when he spent 30 hours searching for a specific couch, an experience shared by many online shoppers. The duo identified significant inefficiencies and trust issues in existing e-commerce search engines, which often resulted in shoppers spending extensive time across multiple websites to make a purchase over $50.
Alex and Zach describe their journey from basic vector search techniques to developing advanced AI integrations, including natural language processing and color data sets, to enhance search capabilities.
They introduce their self-learning system that improves search results based on user queries, utilizing the principle of compositionality. This innovative approach allows their search engine to generate outputs far beyond its training set, facilitated by a highly expressive knowledge graph database.
The conversation explores philosophical influences like Douglas Hofstadter and symbolic logic, advocating for a neuro-symbolic AI system as the next frontier.
They highlight the recent shift towards consumer AI driven by genuine pain points and the novelty of AI solutions, making such investments appealing and less risky. Onton's success in reducing search friction and enhancing personalization demonstrates the value of consumer AI.
Addressing the long-term defensibility of Onton, they emphasize their unique, clean data pipeline and search index. They also discuss the necessity of solving current AI limitations to progress toward AGI, expressing skepticism about immediate breakthroughs.
Guest Links:
Rob Links
01:00 Co-founders solved e-commerce search problem differently.
04:52 Developed robust parser for natural language processing.
08:04 Shopping is time-consuming and depends on data accuracy.
12:33 Blended symbolic and connectionist approaches enhance AI.
15:00 AI drives renewed interest in consumer startups profitability.
19:12 Core business products are defensible, not minor improvements.
21:29 AGI's actions depend on differing objective functions.
23:34 Finding value in small moments amid life's anxieties.
52 episodes
Share
