AI Agents Podcast Building AI Agents That Learn - Insights from Elman Mansimov, Founder of Sourcely | EP63
24 snips
Jul 31, 2025 Elman Mansimov, a research scientist at DeepMind and founder of Sourcely, discusses his innovative journey in AI. He shares insights on building intelligent agents capable of learning and adapting. Elman emphasizes the role of context and reasoning in enhancing AI tools, while explaining the challenges faced in developing Sourcely for deep search and source verification. The conversation also touches on the evolving landscape of AI in academia and industry, and the need for human creativity amidst rapid technological advancement.
AI Snips
Chapters
Transcript
Episode notes
Sourcely's Origin Story
- Elman Mansimov shared how Sourcely started as a tool by UK dental students to quickly find paper sources.
- After acquisition, he expanded its use beyond academia to fields like fitness and nutrition for evidence verification.
AI's Learning Advantage
- Traditional programming relies on algorithms and rules, but machine learning enables computers to learn patterns from data.
- Neural networks helped AI advance by learning hierarchical features, improving tasks like image recognition.
Boost from AI Coding Tools
- Elman described how Copilot supercharged his coding by helping with unfamiliar web tech.
- He combined tools like ChatGPT with Copilot to learn and solve coding challenges efficiently and enjoyably.