

338: From Extraction to Understanding: Martin Goodson, CEO of Evolution AI, on Why AGI Is The Wrong Goal
Jun 2, 2025
Martin Goodson, CEO of Evolution AI and former Oxford researcher, shares his fascinating journey from biological data to pioneering AI in optical character recognition. He discusses the shortcomings of traditional OCR and how contextual understanding transforms document intelligence. The conversation touches on the importance of domain knowledge in financial analysis and the challenges of accuracy in AI, particularly with numerical data. Goodson questions the pursuit of AGI, advocating for AI systems that enhance human capabilities rather than mimic them.
AI Snips
Chapters
Books
Transcript
Episode notes
Beyond OCR: Document Understanding
- Combining visual and linguistic context unlocks deeper document understanding beyond OCR.
- Large language models can integrate cross-domain information and background knowledge, improving financial analysis.
Complementary Strengths of OCR and LLMs
- OCR is precise in extracting characters but lacks true document understanding.
- Large language models have understanding but are prone to hallucinations, requiring combined techniques for accuracy.
Challenges with Numbers in LLMs
- Numbers pose a greater challenge than language for LLMs, leading to errors and hallucinations.
- LLMs understand financial language well but struggle with precise numerical accuracy.