7 Questions with Jeremy Howard (Answer.ai, fast.ai) on Open Source AI and Agents

Value of the Information & Strength of the Argument

The interview provides valuable insights from a prominent AI figure. Howard’s arguments are well-reasoned, drawing on historical analogies (e.g., Microsoft’s stance on open source) and practical experience. He offers a clear perspective on the risks of over-automation and the importance of human learning. The discussion of Solveit and its applications in various fields adds practical value. His method for evaluating AI hype—reading papers and experimenting—is pragmatic and credible.

Scientific Rigor, Source Quality, Title Accuracy

The interview is based on Howard’s expertise and opinions; no formal sources are cited. He references specific projects (e.g., Cartridges, DeepSeek, Parakeet) and historical events, but without detailed citations. The title accurately reflects the content. The discussion is coherent and aligns with Howard’s public advocacy for open source and human-centered AI.

The title accurately reflects the content: a series of seven questions covering open source AI, agents, and related topics.

8/10

Jeremy Howard is a well-known AI researcher and educator, co-founder of fast.ai and Answer.ai. The interview is based on his experience and opinions, not on new empirical data. He references specific projects and papers, but no formal citations are provided. The content is coherent and aligns with his public positions.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 00:30Howard expresses disappointment that PyTorch community focuses on agents, contrary to PyTorch's human-centric philosophy.
• 02:30He warns that going all-in on agents could lead to obsolescence and organizational decay.
• 04:00Introduction of Solveit, a method for using AI to support human learning and agency.
• 06:00Application of Solveit to non-coding fields like writing, business strategy, and project management.
• 08:00Howard explains how foundational knowledge helps him separate hype from real advances.
• 10:00He highlights recent techniques: hybrid retrieval, diffusion models, and NVIDIA's Parakeet.
• 12:00Defense of open source AI on economic and safety grounds, citing historical success.
• 14:00Reflection on fast.ai's mission to democratize AI, and receiving a DGX Spark from NVIDIA.

• PyTorch Conference 2025 — Event where the interview took place.
• Solveit — Method developed by Answer.ai for human-centered AI assistance.
• George Polya's book — Inspiration for Solveit's four-step problem-solving process.
• Cartridges (Stanford) — Research on external memory and retrieval.
• DeepSeek — Recent work on OCR as a different encoding.
• NVIDIA Parakeet — Speech-to-text model praised for efficiency.

• Open Source Initiative — Supports the value and safety of open source software.

• AI safety concerns with open source — Some argue that open source AI models can be misused, a point Howard counters.