We're sunsetting PodQuest on 2025-07-28. Thank you for your support!
Export Podcast Subscriptions
cover of episode 338: From Extraction to Understanding: Martin Goodson, CEO of Evolution AI, on Why AGI Is The Wrong Goal

338: From Extraction to Understanding: Martin Goodson, CEO of Evolution AI, on Why AGI Is The Wrong Goal

2025/6/2
logo of podcast AI and the Future of Work

AI and the Future of Work

AI Deep Dive AI Chapters Transcript
People
D
Dan Turchin
M
Martin Goodson
Topics
Martin Goodson: 我认为语言并非智能的首要方面。我在成为父亲后发现,孩子们在掌握语言之前就已经表现出智能行为。例如,我的女儿在还不会说话的时候,就通过模仿撞头来告诉我她发生了什么。我们是否过于强调语言在智能中的作用?现在的机器在语言方面表现出色,但在其他方面却存在不足。我们是否应该重新思考人工智能的目标,不应仅仅是模仿人类智能,而是应该关注更广泛的智能形式。 Dan Turchin: 我赞同Martin的观点,我们应该追求碳基生物和硅基机器之间最佳的合作关系。机器擅长处理人类不擅长的单调乏味的任务,例如基于数百万页文档来确定贷款规模。我们应该思考如何让机器更好地补充人类,而不是仅仅追求通过图灵测试,让人类误以为自己在与人类互动。

Deep Dive

Shownotes Transcript

Dr. Martin Goodson is the founder and CEO of Evolution AI, a company he launched in 2012 to apply deep learning to optical character recognition (OCR). The company has received one of the largest AI R&D grants ever awarded by the UK government, along with investment from First Minute Capital. A former scientific researcher at Oxford University, Martin has led AI research across several organizations and was elected Chair of the Data Science and AI Section of the Royal Statistical Society in 2019.

In this conversation, we discuss:

  • Martin Goodson’s journey from researching biological data to founding Evolution AI and pioneering deep learning for document understanding.
  • Why traditional OCR missed the mark, and how combining visual and linguistic context unlocked a new frontier in document intelligence.
  • The evolution from data extraction to true financial analysis, and why domain knowledge is essential for reading statements like income reports.
  • The risks of LLM hallucinations, especially with numerical data, and why accuracy still requires combining techniques across model types.
  • What Martin believes intelligence really is, and why language alone may be the wrong benchmark for AGI.
  • Why recreating human intelligence shouldn’t be the goal of AI research, and how we can build systems that support, not mimic, human thinking.

Resources:

  • Subscribe to the AI & The Future of Work Newsletter)
  • Connect with Martin on LinkedIn)
  • Check out the YouTube channel of the London Machine Learning Meetup)
  • AI fun fact article)
  • On How to Ovecome Imposter Syndrome)
  • Past episodes mentioned:
  • On Why doing Taxes is like finding the Best Route on a Map) with Daniel Marcous
  • On Making AI Smarter Without Harming Humans) with Peter Voss